Fix AttributeError on SignalR disconnect during HA shutdown

[njloof]

Summary

SignalRHub.disconnect() called self._client.stop(), but pysignalr.SignalRClient (>=1.3.0) has no stop() or close() method. Its run() coroutine blocks until cancelled and is designed to be stopped by cancelling the task that's awaiting it (see pysignalr's WebsocketTransport.run() — an unconditional reconnect loop with no external stop hook).

This raised AttributeError: 'SignalRClient' object has no attribute 'stop' on every graceful Home Assistant shutdown/restart (fired from the EVENT_HOMEASSISTANT_STOP listener in custom_components/hilo), which then cascaded into a follow-on RuntimeError: aclose(): asynchronous generator is already running from the interrupted cleanup.

Impact

Limited to shutdown. It did not prevent Home Assistant from stopping and did not affect the next startup (a fresh SignalRClient is always created on the next run()). Cosmetic but real — it flooded the logs with an ERROR + traceback on every single restart.

Fix

SignalRHub now records the task running run() via asyncio.current_task(). disconnect() cancels and awaits that task instead of calling the nonexistent client method.

Testing

Added tests/test_signalr.py:

• test_disconnect_cancels_running_task: starts run() in a task, confirms disconnect() cancels it cleanly without raising.
• test_disconnect_without_a_running_task_is_a_no_op: confirms calling disconnect() before run() has ever started doesn't raise.

Confirmed the new test fails with the original AttributeError against the pre-fix code (git stash + rerun) and passes against the fix.

Full suite: 66 passed. ruff check/ruff format, mypy, bandit, and codespell all clean on the changed files.

[ic-dev21]

I see you found the same thing I did while exploring pysignalR's library lol