fix: add timeout to mqtt connect to avoid indefinite hang

[garymm]

Summary

get_rest_devices waits on the MQTT connect future with no timeout:

connect_future = await loop.run_in_executor(None, mqtt_connection.connect)
await loop.run_in_executor(None, connect_future.result)

If the Hatch cloud is unreachable in a way that doesn't fail fast (hung TCP, half-open connection during an outage), connect_future.result blocks its executor thread and the awaiting coroutine forever. In the ha_hatch integration this manifests as config entry setup hanging indefinitely during Hatch cloud outages.

This applies the same fix as #60 did for the shadow client operations: a bounded wait (30 s) that raises concurrent.futures.TimeoutError, which propagates through the existing except block in get_rest_devices so callers can retry with backoff.

Context

During a Hatch cloud outage on 2026-07-24 this hang (together with a companion bug in ha_hatch, PR incoming there) repeatedly locked up a Home Assistant instance from ~4:41 AM until the integration was disabled.

Testing

• pytest tests/ — 23 passed

🤖 Generated with Claude Code

https://claude.ai/code/session_01KUv54T1tjc6TJYCsUXyPE2

[yoniamir]

I independently observed the connect-side failure today on a separate production Home Assistant instance using hatch_rest_api 1.32.0. During repeated Core startups, get_rest_devices logged AWS_ERROR_MQTT_UNEXPECTED_HANGUP; Core then stopped answering its HTTP and WebSocket APIs. Safe mode restored service immediately, and disabling only the Hatch config entry allowed a normal stable restart.

A bounded Future.result(timeout=...) is the correct direction: failure of the Hatch cloud must make the integration unavailable, never leave config-entry setup pending indefinitely.

Please add regression tests covering:

1. a mocked connect future that never resolves and raises TimeoutError within the configured bound;
2. proof that the asyncio event loop remains responsive while the executor wait is pending; and
3. propagation of a connect future that completes with an AWS exception, preserving the existing retry/error behavior.

It would also be useful to confirm that a timed-out native connect attempt is cleaned up or cannot accumulate across retries. Companion Home Assistant event-loop fix: dahlb/ha_hatch#320.