Handle cancelation in wait_for_ble_connections_free
If `wait_for_ble_connections_free` was canceled due to timeout or
the esp disconnecting from Home Assistant the future would get
canceled. When we reconnect and get the next callback we need
to handle it being done.
fixes
```
2023-03-21 02:34:36.876 ERROR (MainThread) [homeassistant] Error doing job: Fatal error: protocol.data_received() call failed.
Traceback (most recent call last):
File "/usr/local/lib/python3.10/asyncio/selector_events.py", line 868, in _read_ready__data_received
self._protocol.data_received(data)
File "/usr/local/lib/python3.10/site-packages/aioesphomeapi/_frame_helper.py", line 195, in data_received
self._callback_packet(msg_type_int, bytes(packet_data))
File "/usr/local/lib/python3.10/site-packages/aioesphomeapi/_frame_helper.py", line 110, in _callback_packet
self._on_pkt(Packet(type_, data))
File "/usr/local/lib/python3.10/site-packages/aioesphomeapi/connection.py", line 688, in _process_packet
handler(msg)
File "/usr/local/lib/python3.10/site-packages/aioesphomeapi/client.py", line 482, in on_msg
on_bluetooth_connections_free_update(resp.free, resp.limit)
File "/usr/src/homeassistant/homeassistant/components/esphome/entry_data.py", line 136, in async_update_ble_connection_limits
fut.set_result(free)
asyncio.exceptions.InvalidStateError: invalid state
```
* Add auto repairs for events schema
* Add auto repairs for events schema
* Add auto repairs for events schema
* Add auto repairs for events schema
* Add auto repairs for events schema
* fix bug - wrong table
* Fix index not being dropped on postgresql databases with a schema prefix
Added logging in case index drops fail so we can tell
why in the future
* coverage
* Bump PyJWT to 2.6.0
* fix time being frozen too late which makes the access token creation time in the future
* revert zha change
* fix repairs test
* fix ical test
* Rename `zha.api` to `zha.websocket_api`
* Implement a ZHA network settings API
* Use the enum name as the radio type
* Don't filter out ignored config entries
* [WIP] Start unit tests
* Add unit tests
* Rename ZHA websocket API module in `.coveragerc`
* Rename `api` to `websocket_api`
* Increase test coverage to 100%
* Allow passing binary to the WS connection
* Expand test coverage
* Test non-existing handler
* Allow signaling end of stream using empty payloads
* Store handlers in a list
* Handle binary handlers raising exceptions
If the time period for the mean/time weighted average was smaller
than we can measure (less than one microsecond), generating
statistics would fail with a divide by zero error. This is likely
only happens if the database schema precision is incorrect.
* Add text to speech and stages to pipeline
* Default to "cloud" TTS when engine is None
* Refactor pipeline request to split text/audio
* Refactor with PipelineRun
* Generate pipeline from language
* Clean up
* Restore TTS code
* Add audio pipeline test
* Clean TTS cache in test
* Clean up tests and pipeline base class
* Stop pylint and pytest magics from fighting
* Include mock_get_cache_files
* Break out statistics schema repairs into a repairs module
A future PR will add repairs for events, states, etc
* reorg
* reorg
* reorg
* reorg
* fixes
* fix patch targets
* name space rename
