Release 2.8.0: correctness, robustness and accuracy fixes

[kmich]

Description

Fixes from a full code review, released as 2.8.0. The full list is in CHANGELOG.md.

Confirmed bugs

• Stuck-sensor detection never fired, because its state lived on a dict rebuilt every cycle. It is now time-based: temperature is flagged after 4 h without moving, humidity and pressure after 6 h. Humidity near 0 % or 100 % is excluded.
• HDD/CDD season totals: the season received the new day's first reading instead of the completed day. The CDD season also never reset, and a day that ended while Home Assistant was stopped was lost.
• Service crash: reset_rain_baseline, reset_learning_state and export_learning_state raised AttributeError when called without entry_id. The event platform had stored a plain dict in hass.data[DOMAIN]; event entities now live on the coordinator.
• FWI: the moisture codes got a second drying step on every intraday recompute, and the daily update ran at midnight. They now advance once, at local noon, and ISI/BUI/FWI/DSR are recomputed from the stored codes (new fwi_indices() helper).
• Pressure-trend window: the configured window reverted to 3 h after a restart.
• Irrigation score: it read rain-today before that value was set, so rain never offset ET₀.
• WU/PWSWeather: dailyrainin was the rolling 24 h total; it is now rain since local midnight.
• Hourly forecast: local readings were blended twice (coordinator + weather entity). Only the weather entity's time-aware blend remains.
• Hourly ET₀ now follows local solar time instead of assuming 06–18 UTC. Hargreaves and Penman-Monteith use the FAO-56 daily mean temperature.
• Vigicrues river flow was 1000× too high: Hub'Eau reports discharge (Q) in L/s.

Robustness and security

• Background loops go through _start_periodic / _start_once. A failed run is logged and retried, and every task is cancelled on unload.
• Repairs issue IDs are now per config entry. Legacy shared IDs are cleaned up, and an entry's issues are removed with it (async_remove_entry).
• API keys can no longer reach the log. OWM and Pirate Weather no longer raise the URL-bearing ClientResponseError, and error messages are scrubbed.
• The coordinator and config flow use HA's shared HTTP session.
• The weather entity pushes forecast updates to subscribed cards.

Accuracy

• Frost point: water-referenced humidity, inverted on the ice curve.
• US AQI: 2024 EPA PM2.5 breakpoints, with EPA truncation.
• The wind-speed calibration offset now also applies to gust.
• kPa, psi and ft/s source units are converted.
• A 0 lx reading is kept as-is instead of being treated as missing.
• CWOP sends h00 for 100 % humidity and adds the since-midnight rain field P.

Related Issue

None. These came out of a repository-wide code review rather than a reported issue.

Motivation and Context

Several derived sensors were silently wrong (degree-day seasons, FWI, river flow, irrigation score, stuck detection), and three services crashed. A few changes alter sensor values users may compare against thresholds:

• FWI will read higher during the day.
• PM2.5 between 9 and 12 µg/m³ is now "Moderate" instead of "Good".
• River flow drops by a factor of 1000.
• Stuck-sensor Repairs issues can now appear, one per station.

How Has This Been Tested?

• pytest tests/ on Python 3.12 with Home Assistant 2024.12.5: 420 passed. This includes 27 new regression tests in tests/test_review_fixes.py and tests/test_uploaders.py covering each fix above.
• ruff check and ruff format --check pass on custom_components/, tests/ and scripts/.
• scripts/validate_dashboard_entities.py passes, and the manifest/pyproject versions match.
• Not tested against a live Home Assistant instance or the live external APIs. The Hub'Eau L/s unit was checked against its documentation.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update

Checklist:

• My code follows the code style of this project
• I have updated the documentation accordingly (if needed)
• I have added tests to cover my changes (if applicable)
• All new and existing tests passed
• I have updated CHANGELOG.md with my changes

Not in this PR: splitting _compute into tick-driven accumulation and idempotent derivation, de-duplicating the config/options flow steps, moving to entry.runtime_data, using HA-native unit conversion, and the CI workflow duplication.