fix: retry sooner on transient DNS failures during device update

[mfncl9991]

Summary

During live testing of PR #158, occasional transient DNS failures were observed:

HomeAssistantError: Error communicating with Winix: Cannot connect to host us.api.winix-iot.com:443 ssl:default [Timeout while contacting DNS servers]

DNS recovers on its own, but with no retry logic the coordinator waits the full scan interval (30s) before trying again, leaving entities unavailable in the meantime.

Change

Catch ClientConnectorDNSError in _async_update_data and raise UpdateFailed(retry_after=timedelta(seconds=15)), letting the DataUpdateCoordinator framework retry sooner rather than waiting the full interval.

ClientConnectorDNSError is already wrapped into HomeAssistantError by the driver, so it's detected via __cause__. All other HomeAssistantError cases (HTTP errors, timeouts) are re-raised unchanged.

try:
    for device_wrapper in self._device_wrappers:
        await device_wrapper.update()
except HomeAssistantError as err:
    if isinstance(err.__cause__, aiohttp.ClientConnectorDNSError):
        raise UpdateFailed(retry_after=timedelta(seconds=15)) from err
    raise

References

• Discussed in PR fix: update auth and device control for Winix API v1.5.7 #158 comments

[iprak]

Thanks for looking into this and getting to it before me.