fix: clean up orphaned .sql on compression failure and fix stale startup log

- backup-lib: delete uncompressed .sql file in catch block when gzip
  compression fails, preventing silent disk usage accumulation
- server: replace stale retentionDays scalar with retentionSource in
  startup log since retention is now read from DB on each backup tick

packages/db/src/backup-lib.ts

@@ -763,6 +763,9 @@ export async function runDatabaseBackup(opts: RunDatabaseBackupOptions): Promise
     if (existsSync(backupFile)) {
       try { unlinkSync(backupFile); } catch { /* ignore */ }
     }
+    if (existsSync(sqlFile)) {
+      try { unlinkSync(sqlFile); } catch { /* ignore */ }
+    }
     throw error;
   } finally {
     await sql.end();

server/src/index.ts

@@ -670,7 +670,7 @@ export async function startServer(): Promise<StartedServer> {
     logger.info(
       {
         intervalMinutes: config.databaseBackupIntervalMinutes,
-        retentionDays: config.databaseBackupRetentionDays,
+        retentionSource: "instance-settings-db",
         backupDir: config.databaseBackupDir,
       },
       "Automatic database backups enabled",