Keep API available when MQTT fails

[MTrab]

Summary

Keep the cloud API usable when MQTT cannot connect, expose MQTT connection state, and reduce aggressive MQTT reconnect behavior that could contribute to broker rate limits.

Changes

This keeps authentication and API refresh working even when the initial MQTT connection fails, then retries MQTT quietly in the background. MQTT reconnect timing is slowed down, paho reconnect backoff is increased, and token refresh no longer forces immediate MQTT reconnects while disconnected.

The MQTT connection state is exposed on both the cloud object and devices, MQTT commands now fail with a clear connection error when MQTT is unavailable, and paho reconnect callbacks resubscribe existing topics so recovered connections remain usable. Legacy unused MQTT resume code and unused private state were removed.

Testing

• git status: clean before PR creation
• poetry run ruff format pyworxcloud tests
• poetry run ruff check pyworxcloud tests
• poetry run pytest tests/test_mqtt_runtime.py tests/test_mqtt_lifecycle.py tests/test_mqtt_commands.py tests/test_api_lifecycle.py tests/test_connect_cleanup.py
• poetry run pytest

Notes

Proposed semver label: patch. This label has not been set yet and should be explicitly approved before merge.