One device per scene playlist (model "Scene Playlist"), each with:
- a switch (start/stop; the server cycles one playlist at a time, so starting one stops any other)
- a "Current Scene" sensor that resolves the active preset name from scene_presets while this playlist is cycling
- an "Items" diagnostic sensor (configured scene count; no state_class)
The coordinator reads scene_playlists plus the flat playlist_state from the /api/v1/snapshot payload and gains start_playlist()/stop_playlist(); __init__ registers and prunes per-playlist devices and reloads on playlist-id changes; the event listener also refreshes on the playlist_state_changed WS event. Shared device/lookup/running-state plumbing lives in a new LedGrabPlaylistEntity base (entity.py) used by both the switch and the sensors. Adds en/ru translation keys.
Note: this also lands the in-progress coordinator migration from the per-request fan-out to the single /api/v1/snapshot poll that was already present in the working tree.
Requires the led-grab server /api/v1/snapshot to emit scene_playlists + playlist_state (companion server change, tracked separately).
alexei.dolgolyov
705616f8b0