Fix ClearLogsWorker never running and remove unsafe VACUUM

Three bugs:
- ExistingPeriodicWorkPolicy.REPLACE was resetting the 5-minute initial
  delay on every app start, so the worker never fired for users who open
  the app more than once before the delay expires. Changed to KEEP.
- No immediate cleanup meant existing accumulated data (e.g. 4GB) sat
  untouched until the periodic worker eventually ran. Added a one-time
  WorkRequest (ExistingWorkPolicy.KEEP) on every app start so cleanup
  runs right away.
- VACUUM on a multi-GB SQLite database requires equivalent free disk
  space as temp storage and blocks the DB for a long time — removed it.
  Row-level deletion via count/age caps is sufficient to control size.

https://claude.ai/code/session_01BCMhs2yA9XNUvfaX4sqsfv

app/src/main/java/com/greenart7c3/nostrsigner/Amber.kt

@@ -14,6 +14,8 @@ import androidx.lifecycle.LifecycleObserver
 import androidx.lifecycle.LifecycleOwner
 import androidx.lifecycle.ProcessLifecycleOwner
 import androidx.work.ExistingPeriodicWorkPolicy
+import androidx.work.ExistingWorkPolicy
+import androidx.work.OneTimeWorkRequestBuilder
 import androidx.work.PeriodicWorkRequestBuilder
 import androidx.work.WorkManager
 import coil3.EventListener
@@ -207,18 +209,31 @@ class Amber :
     }
 
     private fun startCleanLogsAlarm() {
-        val workRequest = PeriodicWorkRequestBuilder<ClearLogsWorker>(
+        val workManager = WorkManager.getInstance(this)
+
+        // Run an immediate one-time cleanup on every startup to handle accumulated data.
+        // KEEP policy ensures only one instance runs at a time if the app is restarted quickly.
+        val immediateRequest = OneTimeWorkRequestBuilder<ClearLogsWorker>()
+            .addTag("clearLogsWorkOneTime")
+            .build()
+        workManager.enqueueUniqueWork(
+            "ClearLogsWorkerOneTime",
+            ExistingWorkPolicy.KEEP,
+            immediateRequest,
+        )
+
+        // Also schedule a daily periodic run.
+        // KEEP policy preserves the existing schedule so app restarts don't reset the timer.
+        val periodicRequest = PeriodicWorkRequestBuilder<ClearLogsWorker>(
             24,
             TimeUnit.HOURS,
         )
-            .setInitialDelay(5, TimeUnit.MINUTES) // Delay first run by 5 minutes
             .addTag("clearLogsWork")
             .build()
-
-        WorkManager.getInstance(this).enqueueUniquePeriodicWork(
+        workManager.enqueueUniquePeriodicWork(
             "ClearLogsWorker",
-            ExistingPeriodicWorkPolicy.REPLACE,
-            workRequest,
+            ExistingPeriodicWorkPolicy.KEEP,
+            periodicRequest,
         )
     }
 

app/src/main/java/com/greenart7c3/nostrsigner/service/ClearLogsWorker.kt

@@ -47,10 +47,6 @@ class ClearLogsWorker(appContext: Context, workerParams: WorkerParameters) : Cor
                     Log.d(Amber.TAG, "Trimmed $excessLogs excess log entries (cap: $MAX_LOG_ENTRIES)")
                 }
 
-                // Reclaim freed space from SQLite pages
-                logDatabase.openHelper.writableDatabase.execSQL("VACUUM")
-                historyDatabase.openHelper.writableDatabase.execSQL("VACUUM")
-
                 val database = Amber.instance.getDatabase(it.npub)
                 database.dao().updateExpiredPermissions(TimeUtils.now())
                 val deleted = database.dao().deleteOldApplications(now / 1000)