mobile: instant reattach on foreground network change (Wi-Fi↔cellular)

[pallaoro]

Summary

Make mobile auto-reattach instant on a network change that happens while the app stays in the foreground (e.g. Wi-Fi ↔ cellular, without backgrounding).

Background

Auto-reattach shipped in #70 (apps/mobile/App.tsx, apps/mobile/src/connection.ts). It covers the two dominant triggers instantly:

• App foreground — AppState active → probe (Connection.ping()) → reconnect if dead.
• Socket drop — the transport's onClose → reconnect.

Both back off (1s → 15s) and rebind the board + open terminal to the fresh api.

The gap

A network change while the app stays foregrounded isn't caught instantly. We rely on the socket's onClose + the 15s keepalive (connection.ts) to notice. If the flip leaves a half-open socket that never fires a close event, reattach can lag up to ~15s (until the next keepalive ping fails).

This is a secondary case — the keepalive catches it eventually, and the far more common "app-switch" path is already instant via AppState. Filing so it's tracked, not lost.

Proposed fix

Add a network-state change listener that triggers the existing reconnect() immediately when connectivity returns.

• expo-network exposes Network.addNetworkStateChangeListener — confirmed not currently installed (apps/mobile/node_modules has neither expo-network nor @react-native-community/netinfo).
• Cost: a native dependency → requires an expo prebuild + pod install cycle. That's the reason it was deliberately deferred rather than bundled into would it be possible to create #70 (which added zero native deps).

Acceptance

• Toggling Wi-Fi/airplane mode off→on while a task terminal is open reattaches within ~1–2s (not up to 15s).
• No new native module beyond the network listener; reuses the existing reconnect() path.

Priority

Low. Add only if the ~15s worst-case lag is felt in practice. See the Clawnify decision doc "Ateam Mobile Connectivity — Decision: SSH-to-any-host vs Tailscale."