fix(audio): stop emitting a transient idle during a source change

Root cause of the disappearing media notification, and it is self-inflicted
on EVERY Android version — no plugin patch involved.

`audio_service`'s `_observePlaybackState` (audio_service.dart:1131-1136) calls
`AudioService._stop()` on ANY transition into `idle` from a non-idle state.
That reaches `stopService()` -> `deactivateMediaSession()` ->
`notificationManager.cancel(NOTIFICATION_ID)`. The notification is re-posted
at exactly one place, `internalStartForeground()`, reachable only from the
`!wasPlaying && playing` edge in `setState()`, and its FIRST statement is
`ContextCompat.startForegroundService(...)` — which on API 31+ throws
`ForegroundServiceStartNotAllowedException` whenever the process is not in a
foreground state.

Every station change walked straight into that. `_cambiarFuente` pushes
`loading`, then `_recrearPlayer` disposes the old `AudioPlayer` and builds a
FRESH one; a fresh player's first `playerStateStream` event is always `idle`,
and the listener forwarded it verbatim. So `loading -> idle` tore the
foreground service down mid-source-change, and recovery depended on the
following `playing: true` edge restarting it. Screen off, lock screen, or an
Android Auto / Bluetooth-initiated start is precisely where the platform
refuses that restart: audio keeps playing, the notification never returns.
That is exactly what the user reports.

The mapping decision moves out of the private `_mapProcState` into a pure
top-level `mapearEstadoProceso(proc, {required bool cambiandoFuente})`, so the
one line that decides whether the foreground service dies is unit-testable
without instantiating the handler (which needs MethodChannels). It is
byte-for-byte identical to the old switch in every case except `idle` while a
source change is in flight, which now maps to `loading`. The test asserts the
full ProcessingState x cambiandoFuente matrix against a literal transcription
of the previous mapping, and asserts both directions explicitly: a real stop
still yields `idle`, a source-change idle yields `loading`, and `idle` is the
only case where the two branches differ at all.

The only risk this introduces is a `_cambiandoFuente` stuck at `true`: a real
user stop would be masked away from `idle`, the service would never stop, and
the notification would become unkillable. So the flag is cleared by four
independent mechanisms rather than one audited path:

  - a `finally` around the whole body of `_cambiarFuente`, which covers normal
    completion, BOTH `revision != _revisionFuente` early returns, every
    `rethrow` out of a catch clause, and any non-`Exception` `Error` that none
    of the three clauses matches;
  - eagerly at the top of each of the three catch clauses — needed on top of
    the `finally` because `_gestionarErrorReproduccion` calls `_player.stop()`
    WITHOUT awaiting it, so that `idle` could otherwise land while the mask
    was still up;
  - right after `setUrl` resolves, before anything below can await, since the
    fresh player's transient `idle` is already behind us at that point;
  - at the start of `stop()` — before `_player.stop()` — and at the start of
    `_gestionarErrorReproduccion`, which makes the invariant total: the flag
    is `false` before every single `_player.stop()` call in this class.

`stop()` matters most: `BaseAudioHandler.stop()` is empty, so the handler
never pushes `idle` itself — teardown is driven entirely by the player's
emission. A stop landing while a station change was still in flight would
otherwise be masked and the notification would survive the stop.

Audited: two `_player.stop()` call sites exist and both are preceded by a
clear; `_recrearPlayer` has exactly one caller and it is guarded; the old
player cannot emit during `_recrearPlayer` because its subscriptions are
cancelled first.

lib/servicios/servicio_audio.dart

@@ -120,6 +120,57 @@ Emisora emisoraDesdeMediaItem(MediaItem mediaItem) {
   );
 }
 
+/// Maps a `just_audio` [ProcessingState] to the `audio_service`
+/// [AudioProcessingState] pushed into `playbackState`. Identical to the
+/// previous private `_mapProcState` in every case EXCEPT one:
+/// [ProcessingState.idle] maps to [AudioProcessingState.loading] while
+/// [cambiandoFuente] is `true`.
+///
+/// Why that single exception exists — this is the media-notification
+/// regression, not a cosmetic tweak:
+///
+/// `audio_service`'s `_observePlaybackState` (`audio_service.dart:1131-1136`)
+/// calls `AudioService._stop()` — which reaches `stopService()` and cancels
+/// the notification through `deactivateMediaSession()` — on ANY transition
+/// into `idle` from a non-idle state. The notification is posted at exactly
+/// one place, `internalStartForeground()`, reachable only from the
+/// `!wasPlaying && playing` edge, and its FIRST statement is
+/// `ContextCompat.startForegroundService(...)`, which throws
+/// `ForegroundServiceStartNotAllowedException` on API 31+ whenever the
+/// process is not in a foreground state.
+///
+/// Every station change walked straight into that: `_cambiarFuente` pushes
+/// `loading`, then `_recrearPlayer` disposes the old [AudioPlayer] and builds
+/// a FRESH one, and a fresh player's first `playerStateStream` event is
+/// always `idle`. Forwarded verbatim, that is a `loading -> idle` transition,
+/// so the foreground service was torn down mid-source-change and the app then
+/// depended on the following `playing: true` edge to restart it. With the
+/// screen off, on the lock screen, or on an Android Auto / Bluetooth-initiated
+/// start, that restart is exactly the case the platform refuses — audio keeps
+/// playing, the notification never comes back. Self-inflicted, on every API
+/// level, no plugin patch needed: just stop emitting the transient `idle`.
+///
+/// A genuine user stop is unaffected: `stop()` clears the flag BEFORE
+/// `_player.stop()`, so its `idle` still reaches `playbackState` as a real
+/// `idle` and still tears the service down. Pure — no handler dependency — so
+/// the full [ProcessingState] x [cambiandoFuente] matrix is unit-testable
+/// directly.
+AudioProcessingState mapearEstadoProceso(
+  ProcessingState proc, {
+  required bool cambiandoFuente,
+}) {
+  if (cambiandoFuente && proc == ProcessingState.idle) {
+    return AudioProcessingState.loading;
+  }
+  return switch (proc) {
+    ProcessingState.idle => AudioProcessingState.idle,
+    ProcessingState.loading => AudioProcessingState.loading,
+    ProcessingState.buffering => AudioProcessingState.buffering,
+    ProcessingState.ready => AudioProcessingState.ready,
+    ProcessingState.completed => AudioProcessingState.completed,
+  };
+}
+
 /// Custom-action names for the equalizer's `PlaybackStateCompat` custom
 /// actions on the now-playing screen (Design "EQ custom actions", item 4).
 /// Public consts so tests and this file's own `customAction` dispatch share
@@ -416,6 +467,24 @@ class PluriWaveAudioHandler extends BaseAudioHandler
   Future<void> _colaCambioFuente = Future<void>.value();
   int _revisionFuente = 0;
 
+  /// `true` only for the window inside [_cambiarFuente] where the OLD player
+  /// has been disposed and the FRESH one has not loaded its URL yet — the
+  /// window in which `playerStateStream` unavoidably emits a transient
+  /// `idle` that is NOT a stop. [mapearEstadoProceso] masks that one `idle`
+  /// as `loading` so `audio_service` does not tear the foreground service
+  /// (and with it the media notification) down mid-source-change; see that
+  /// function's doc for the full mechanism.
+  ///
+  /// A value stuck at `true` is the ONLY risk this flag introduces: a real
+  /// user stop would then be masked away from `idle` and the service would
+  /// never stop, leaving an unkillable notification. It is therefore cleared
+  /// by a `finally` in [_cambiarFuente] (which covers normal completion,
+  /// both revision-mismatch `return`s, every `rethrow`, and any non-`Exception`
+  /// `Error` that no catch clause matches), AND eagerly at the top of every
+  /// catch clause, AND at the start of [stop] and [_gestionarErrorReproduccion]
+  /// — i.e. before every single `_player.stop()` call in this class.
+  bool _cambiandoFuente = false;
+
   /// Active local-music queue (Design "single load-bearing invariant"):
   /// `null` means "not local-queue playback" — the ONLY gate the
   /// auto-advance/skip/isolation logic reads. Radio never sets this field.
@@ -522,7 +591,10 @@ class PluriWaveAudioHandler extends BaseAudioHandler
             if (colaActiva) MediaAction.skipToNext,
           },
           androidCompactActionIndices: [colaActiva ? 1 : 0],
-          processingState: _mapProcState(proc),
+          processingState: mapearEstadoProceso(
+            proc,
+            cambiandoFuente: _cambiandoFuente,
+          ),
           playing: playing,
           bufferedPosition: _player.bufferedPosition,
           speed: _player.speed,
@@ -608,6 +680,11 @@ class PluriWaveAudioHandler extends BaseAudioHandler
   /// only retry exhaustion (or non-network errors) falls through to the
   /// existing error path, so the user sees a single error — no spam per retry.
   void _gestionarErrorReproduccion(Object error) {
+    // Terminal-error path also ends in `_player.stop()` below, and it is
+    // reachable from `_eventosSub`'s `onError` WHILE a source change is still
+    // in flight. Dropping the mask here keeps the invariant total: the flag
+    // is `false` before every `_player.stop()` call in this class.
+    _cambiandoFuente = false;
     if (_intentarReconexion(error)) return;
 
     String mensaje;
@@ -740,16 +817,6 @@ class PluriWaveAudioHandler extends BaseAudioHandler
     return _textos.audioErrorCannotPlay;
   }
 
-  AudioProcessingState _mapProcState(ProcessingState state) {
-    return switch (state) {
-      ProcessingState.idle => AudioProcessingState.idle,
-      ProcessingState.loading => AudioProcessingState.loading,
-      ProcessingState.buffering => AudioProcessingState.buffering,
-      ProcessingState.ready => AudioProcessingState.ready,
-      ProcessingState.completed => AudioProcessingState.completed,
-    };
-  }
-
   /// Public entry point for EVERY external play (phone `reproducir`, car
   /// `emisora:`/`grupo:`/`pista:` non-path). ALWAYS clears the
   /// local queue FIRST (Design ADR-2, the single load-bearing invariant:
@@ -889,16 +956,31 @@ class PluriWaveAudioHandler extends BaseAudioHandler
         errorMessage: null,
       ),
     );
+    // Opens the masking window BEFORE `_recrearPlayer`, which is what
+    // disposes the old player and constructs the fresh one whose first
+    // `playerStateStream` event is the transient `idle` we must not forward
+    // (see [mapearEstadoProceso]).
+    _cambiandoFuente = true;
     try {
       await _recrearPlayer();
       if (revision != _revisionFuente) return;
 
       await _player.setUrl(mediaItem.id).timeout(_timeoutCambioFuente);
+      // Source swap complete: the fresh player's transient `idle` is behind
+      // us, so stop masking immediately — before anything below can await —
+      // and let a real `idle` through again from here on.
+      _cambiandoFuente = false;
       if (revision != _revisionFuente) return;
 
       _iniciarPlaySinBloquear(mediaItem, revision);
       unawaited(_activarEcualizador());
     } on PlayerException catch (e) {
+      // Cleared BEFORE `_gestionarErrorReproduccion`, not just by the
+      // `finally`: that method calls `_player.stop()` without awaiting it, so
+      // the resulting `idle` could otherwise land while the mask was still
+      // up and be rewritten to `loading`. Same reason in the two clauses
+      // below.
+      _cambiandoFuente = false;
       if (revision == _revisionFuente) {
         _gestionarErrorReproduccion(e);
         // Reconnect engaged: complete normally so callers do not surface a
@@ -907,6 +989,7 @@ class PluriWaveAudioHandler extends BaseAudioHandler
       }
       throw Exception(_mensajeAmigable(e));
     } on TimeoutException catch (e) {
+      _cambiandoFuente = false;
       // A real network drop usually surfaces as our 12s source timeout:
       // route it through the reconnect machine instead of a terminal error.
       if (revision == _revisionFuente) {
@@ -915,6 +998,7 @@ class PluriWaveAudioHandler extends BaseAudioHandler
       }
       rethrow;
     } on Exception catch (e, stackTrace) {
+      _cambiandoFuente = false;
       developer.log(
         '[PluriWave] Error inesperado en playMediaItem: $e',
         name: 'ServicioAudio',
@@ -933,6 +1017,15 @@ class PluriWaveAudioHandler extends BaseAudioHandler
         this.mediaItem.add(null);
       }
       rethrow;
+    } finally {
+      // Leak-proof backstop. Dart runs `finally` on EVERY exit from the
+      // block above: normal completion, both `revision != _revisionFuente`
+      // early returns, every `throw`/`rethrow` out of a catch clause, and —
+      // crucially — any `Error` (as opposed to `Exception`) that none of the
+      // three clauses catches. The flag must never depend on a single
+      // hand-audited exit path, because a `_cambiandoFuente` stuck at `true`
+      // would mask a REAL stop's `idle` and leave the notification unkillable.
+      _cambiandoFuente = false;
     }
   }
 
@@ -1147,6 +1240,17 @@ class PluriWaveAudioHandler extends BaseAudioHandler
     // logic above — `onTaskRemoved` (which calls stop()) inherits this for
     // free (Phase 3 task 3.5).
     _desactivarCola();
+    // Genuine user stop: drop the source-change mask BEFORE `_player.stop()`,
+    // so its `idle` reaches `playbackState` as a REAL `idle` and
+    // `audio_service` tears the foreground service down as it always has.
+    // `stop()` never pushes `idle` itself — `BaseAudioHandler.stop()` is
+    // empty and the teardown is driven entirely by the player's emission —
+    // so a stop landing while a station change is still in flight would
+    // otherwise be masked to `loading` and the notification would become
+    // unkillable. Paired with `_revisionFuente++` below, which invalidates
+    // that in-flight change; its `finally` only re-clears the flag, and
+    // nothing re-arms it (the single `= true` assignment already ran).
+    _cambiandoFuente = false;
     _revisionFuente++;
     await _player.stop();
     emisoraActual = null;