Fix dirty client state after failed BLE connection

Previously a failed connect() / start_notify() left self._client pointing
at a broken BleakClient. On the next poll is_connected returned False but
the stale object caused confusing errors. Now the client is only assigned
after both steps succeed; on any exception we try a clean disconnect and
reset to None so the next attempt starts completely fresh.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

custom_components/xiaoxiang_bms/bluetooth_handler.py

@@ -52,14 +52,25 @@ class BmsBluetoothHandler:
         if self.is_connected:
             return
         _LOGGER.debug("Connecting to BMS at %s (via %s)", self._address, ble_device.name)
-        self._client = BleakClient(
+        client = BleakClient(
             ble_device,
             disconnected_callback=self._on_disconnect,
         )
-        await self._client.connect()
+        try:
-        await self._client.start_notify(RX_CHAR_UUID, self._on_notify)
+            await client.connect()
+            await client.start_notify(RX_CHAR_UUID, self._on_notify)
+        except Exception:
+            # Ensure a failed connect never leaves a broken client behind —
+            # next poll will start fresh
+            try:
+                await client.disconnect()
+            except Exception:
+                pass
+            self._client = None
+            raise
         # Give the BMS a moment to register the notification subscription
         # before we start sending commands — avoids dropped first response
+        self._client = client
         await asyncio.sleep(0.5)
         _LOGGER.debug("Connected to BMS at %s", self._address)