fix(light): an unknown WHAT does not keep a state restored at startup

[GreenGrassBlueOcean]

Follow-up to #456, found live on the MH200 with a test build (MyHOME v2-phase1-architecture + OWNd#59/#60).

Problem

light.light_74 still showed on. Since the restart, the bus only sent:

17:16:33.475 rx *1*19*74##
17:16:33.519 rx *#1001*74*11*111110111111111111110111##

At startup Home Assistant restored "on", which was the state the WHAT 19 bug had left behind. #456 keeps the last state when an unknown WHAT arrives, so that stale "on" was kept forever.

Change

• A state seen on the bus or set from Home Assistant is still kept against an unknown WHAT, as before.
• A state that was only restored at startup becomes unknown on the first fault report. The warning then says its state is unknown instead of keeping its last state.
• _is_on_restored is set where async_restore_last_state restores on/off. It is cleared wherever a real state is set: bus events, _apply_brightness_state, timed on, and the color-temperature turn-on.

Tests

tests/test_component_light.py:

• a restored "on", then a fault: the state becomes unknown and the warning says so; a later bus "off" is kept against the next fault
• a state set from HA (timed on) is kept against a fault
• the live frames from today (*1*19*74## after a restored "on"): a strict xfail while the installed OWNd reports WHAT 19 as on

Results: 2047 passed and 17 xfailed on the pinned OWNd; 2064 passed against OWNd#59 + #60. ruff and mypy --strict on light.py are clean.

🤖 Generated with Claude Code

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!