v1.6.22

Bug fix

Forecast and weather sensors no longer go unavailable on transient timeouts

When a single slow-coordinator endpoint times out (e.g. /api/4/mixer for weather forecast, /api/4/dailystats/details for irrigation forecast), sensors now retain their last known value instead of becoming unavailable.

The fix passes the previous coordinator data into fetch_all_data — on per-endpoint failure the last good value is reused; on full connection failure (auth error, device unreachable) the coordinator still marks all entities unavailable as expected.