MeshCore: radio liveness watchdog + reset/recovery in the driver

[M0LTE]

Parent: #137. During PoC testing a Companion radio became unresponsive to APP_START until a hard reset (esptool --after hard_reset, i.e. a CP2102 DTR/RTS toggle). The MeshCore driver must detect and recover from this autonomously — required for unattended operation.

• Liveness watchdog: periodic heartbeat (APP_START/DEVICE_QUERY) or a serial-activity timeout to detect a mute/hung companion.
• Recovery: hard-reset the ESP32 over CP2102 DTR/RTS (no buttons), re-open the port, re-APP_START, re-apply radio params + channel config, resume the bearer.
• Escalation/backoff: bounded reset attempts with backoff; raise an alarm/metric if recovery keeps failing; surface link state (healthy / resetting / failed) for observability.
• Harden the open path: a post-open settle + APP_START retry — opening the port can itself reset the board and race the first APP_START (seen in the PoC).

Related: device-control + firmware management API.