feat: pre-build playlist cache and make matching interval configurable

- Pre-build playlist items cache during track matching for instant serving
- Add PreBuildPlaylistItemsCacheAsync() to SpotifyTrackMatchingService
- Combines local Jellyfin tracks + external matched tracks in correct Spotify order
- Saves to both Redis and file cache for persistence across restarts
- Change matching interval from hardcoded 30 minutes to configurable (default: 24 hours)
- Add SPOTIFY_IMPORT_MATCHING_INTERVAL_HOURS environment variable
- Set to 0 to only run once on startup (manual trigger still works)
- Add endpoint usage files to .gitignore
- Update documentation in README and .env.example

Rationale: Spotify playlists like Discover Weekly update once per week,
so running every 24 hours is more than sufficient. Pre-building the cache
eliminates slow 'on the fly' playlist building.

All 225 tests pass.

README.md

@@ -401,11 +401,14 @@ SPOTIFY_IMPORT_PLAYLIST_NAMES=Release Radar,Discover Weekly
    - Caches the list of missing tracks in Redis + file cache
    - Runs automatically on startup (if needed) and every 5 minutes during the sync window
 
-3. **Allstarr Matches Tracks** (2 minutes after startup, then every 30 minutes)
+3. **Allstarr Matches Tracks** (2 minutes after startup, then configurable interval)
    - For each missing track, searches your streaming provider (SquidWTF, Deezer, or Qobuz)
    - Uses fuzzy matching to find the best match (title + artist similarity)
    - Rate-limited to avoid overwhelming the service (150ms delay between searches)
    - Caches matched results for 1 hour
+   - **Pre-builds playlist items cache** for instant serving (no "on the fly" building)
+   - Default interval: 24 hours (configurable via `SPOTIFY_IMPORT_MATCHING_INTERVAL_HOURS`)
+   - Set to 0 to only run once on startup (manual trigger via admin UI still works)
 
 4. **You Open the Playlist in Jellyfin**
    - Allstarr intercepts the request