v3.17.0 - Enhanced Resilience & Fresh Start

Fixes critical issue where daemon could appear healthy while notification
stream silently died, causing hours of missed messages.

- Track all notifications (including receipts)
- Restart if no notifications for 30 minutes (even if pings succeed)
- Detects "healthy daemon, dead notification stream" scenario

- Skip messages older than 5 minutes at startup
- Ensures fresh, responsive start after recovery
- No more churning through hours of backlog

- Fixed 8-hour message outage caused by silent notification stream death
- Daemon was responding to pings but not delivering messages

1. TCP Keepalive (10s) → dead connections
2. Heartbeat (30s) → unresponsive daemon
3. Notification monitoring (30 min) → silent stream death [NEW]
4. Skip old messages (5 min) → fresh start [NEW]
5. RPC reconnect (5 attempts with backoff)
6. Watchdog ping (2 min)
7. Circuit breaker (5 restarts in 5 min)