V3-12: crash reporting (code only)

shouldInitializeCrashReporting() and scrubExtra() are pure, fully unit-tested decision/scrubbing functions wired into sentry_flutter via beforeSend/beforeBreadcrumb. Config.crashReportingEnabled defaults off; the DSN is a compile-time --dart-define, not a preference. RecordingEngine gained an injected onUnexpectedStop callback, firing once when restoreAfterProcessDeath finds a trip still 'recording' at launch -- the process died without a user stop.

Marked partially done: the ticket's real acceptance criteria (a forced crash landing in a real Sentry project from a release build, inspecting a real payload) need a Sentry account and a release build this environment can't produce.

lib/src/recording/recording_engine.dart

@@ -68,11 +68,13 @@ class RecordingEngine {
     LiveTelemetry? liveTelemetry,
     int Function()? clock,
     Future<void> Function()? uploadPending,
+    void Function(String reason)? onUnexpectedStop,
   }) : _repo = repository,
        _source = locationSource,
        _live = liveTelemetry ?? LiveTelemetry.instance,
        _now = clock ?? (() => DateTime.now().millisecondsSinceEpoch),
-       _uploadPending = uploadPending;
+       _uploadPending = uploadPending,
+       _onUnexpectedStop = onUnexpectedStop;
 
   final TripRepository _repo;
   final LocationSource _source;
@@ -84,6 +86,11 @@ class RecordingEngine {
   final Future<void> Function()? _uploadPending;
   Timer? _uploadTimer;
 
+  /// V3-12: injected rather than importing a crash reporter directly, the same reasoning
+  /// as [_uploadPending] -- the engine stays free of any opinion about where a report
+  /// goes, and tests need no Sentry client.
+  final void Function(String reason)? _onUnexpectedStop;
+
   /// Unbounded, exactly like the Kotlin `Channel(UNLIMITED)`. Appending is the only work
   /// done on the fix path.
   final List<TrackPoint> _pending = [];
@@ -240,6 +247,12 @@ class RecordingEngine {
     final trip = await _repo.activeTrip();
     switch (trip?.state) {
       case TripState.recording:
+        // A trip still marked `recording` at launch means the previous process ended
+        // without ever calling stop() or pause() -- a crash, an OS kill, or a location
+        // permission revoked out from under the app. This is the failure V3-12 exists to
+        // surface: the recording stopped, silently, and nobody chose that.
+        _onUnexpectedStop?.call('process death mid-recording');
+
         // Resume into a genuinely *new* segment: the time the process was dead is a real
         // gap in the recording and must render as one.
         //