Add external probe that checks /api/health every 60s. If down >2 min consecutively, page WhatsApp via alert file that heartbeat picks up.

Files changed:

• scripts/mc-watchdog.sh — full rewrite: added flock locking, per-instance temp files, proper python3 heredoc writes, cooldown config, fixed state read/write race conditions
• HEARTBEAT.md — added MC watchdog alert section (every heartbeat check for alert file)

System crontab already had * * * * * entry (pre-existing), so watchdog is live immediately.

Architecture:

1. System crontab runs mc-watchdog.sh * * * * *
2. Probes /api/health, tracks consecutive_failures in /tmp/mc-watchdog-state.json
3. At 2nd consecutive failure: writes state/mc-watchdog-alert.json
4. Heartbeat session polls for alert file → relays [alert|watchdog] to WhatsApp
5. On recovery: watchdog deletes alert file
6. 10min cooldown prevents re-alert spam on extended outages