connections.pybricks: fix leaking tasks if race_disconnect() is canceled

If the `race_disconnect()` method is canceled while waiting for either
of the tasks, there is nothing that ever awaits the tasks, so we get
errors like:

    2025-09-09 21:22:25,946: ERROR: asyncio: Task was destroyed but it is pending!
    task: <Task pending name='race_disconnect: Queue.get' coro=<Queue.get() done, defined at python\3.13.0\Lib\asyncio\queues.py:172> wait_for=<Future cancelled>>
    2025-09-09 21:22:25,947: ERROR: asyncio: Task was destroyed but it is pending!
    task: <Task pending name='Task-21' coro=<Event.wait() done, defined at python\3.13.0\Lib\asyncio\locks.py:200> wait_for=<Future pending cb=[Task.task_wakeup()]>>

To avoid this, cancel the tasks if anything goes wrong while waiting.

Author: dlech

pybricksdev/connections/pybricks.py

@@ -349,10 +349,15 @@ def handle_disconnect(state: ConnectionState):
                 disconnect_event.set()
 
         with self.connection_state_observable.subscribe(handle_disconnect):
-            done, pending = await asyncio.wait(
-                {awaitable_task, disconnect_task},
-                return_when=asyncio.FIRST_COMPLETED,
-            )
+            try:
+                done, pending = await asyncio.wait(
+                    {awaitable_task, disconnect_task},
+                    return_when=asyncio.FIRST_COMPLETED,
+                )
+            except BaseException:
+                awaitable_task.cancel()
+                disconnect_task.cancel()
+                raise
 
             for t in pending:
                 t.cancel()