Force the central signaling relay to reconnect — see notify_force_reconnect.

Response shapes:

• {"status": "requested", "token_available": bool} — a reconnect was kicked off.
• {"status": "skipped", "token_available": bool, "reason": "relay_not_running"} — no reconnect happened because no relay instance exists (daemon started without a token, pre-init, or shutdown). The mobile app's auto-heal loop must NOT wait for a state change in this case (it would hang forever).