Fix SELinux backup execution context and probe source metadata before downtime

update.sh

@@ -255,6 +255,7 @@ if [[ $UP_WORKER == false ]]; then
   # Only enqueue here; result.txt, not systemd-run's exit code, proves success.
   # The launcher must not keep copies of our locks; systemd starts a fresh worker.
   systemd-run --no-block --unit=app-backup-upgrade --collect --property=Type=oneshot \
+    --property=SELinuxContext=system_u:system_r:unconfined_service_t:s0 \
     --property=TimeoutStartSec=24h --property=TimeoutStopSec=5min --property=KillMode=mixed \
     /bin/bash "$UPDATE_ROOT/update.sh" --worker "$UP_STATE" 8>&- 9>&- || {
       # Revoke the queued state while still holding the coordinator lock. Even