fix(aprs): prevent duplicate reporters on repeated service starts

Every ACTION_START - and every null intent delivered by START_STICKY - called
startReporting(), which always constructed a new AprsReporter and overwrote the
field without stopping the old one. AprsReporter owns an independent
SupervisorJob + periodic while(isActive) loop, so every overwritten instance
kept reporting forever and could no longer be reached by ACTION_STOP.

Simulation:
  five ACTION_START events: 5 running reporters, 4 leaked -> fixed: 1 / 0
  START + 3 sticky restarts: 4 running, 3 leaked -> fixed: 1 / 0
  mixed real sequence:      4 running, 3 leaked -> fixed: 1 / 0

At the default 10-minute interval, four leaked reporters send 24 duplicate
position packets per hour and open 24 needless connections; this also amplifies
the connect-time socket leak fixed earlier.

startReporting now returns when the current reporter is active. Config changes
remain correct: AprsCard explicitly sends ACTION_STOP before ACTION_START, so
the old reporter is stopped and nulled before the new configuration starts.

:app:compileReleaseKotlin BUILD SUCCESSFUL.

app/src/main/java/com/rtbishop/look4sat/AprsForegroundService.kt

@@ -64,6 +64,16 @@ class AprsForegroundService : Service() {
     }
 
     private fun startReporting() {
+        // onStartCommand reaches here for every ACTION_START and for the null
+        // intent that START_STICKY delivers on restart. Without this guard each
+        // call built a fresh AprsReporter and overwrote the field, leaving the
+        // previous one running with its own scope and timer: the server then
+        // received one duplicate position report per leaked instance per cycle,
+        // and ACTION_STOP could only ever stop the newest one.
+        reporter?.let { existing ->
+            if (existing.isRunning) return
+            existing.stop()
+        }
         val cfg = AprsStore.loadConfig(this)
         if (!cfg.enabled || cfg.callsign.isBlank()) {
             runCatching {