Fix Radarr discovery: activity log seeding + calibrate threshold

- Stage 1 rewritten to use Emby activity log (playback.stop events)
  with ItemId batch-fetching instead of per-library UserData sort;
  SortBy=DatePlayed requires UserId context and errored server-wide

- Scoring simplified: recency (0-50) + play frequency across all
  users (0-50); CommunityRating is not exposed by Emby Items API
  so TMDB rating scoring moved entirely to Stage 2

- Batch IDs in groups of 100 — 1500+ unique item IDs in a 30-day
  window exceeded GET URL limits on a single request

- Threshold lowered to 50 (from 60); with diverse-genre seeds,
  recommendations rarely appear across multiple seeds so breadth
  score stays at 10/40 — threshold 50 yields ~5 adds per run at
  good quality (rating 7.0+, 1k+ votes)

- HOST1_TMDB_API_KEY moved to sit under Radarr section in
  master_host1.conf where it logically belongs

master.conf

@@ -951,7 +951,7 @@
     LIDARR_DISCOVERY_HISTORY="$DATA_DIR/lidarr_discovery_history.db"
 
 # Radarr discovery shared settings
-    RADARR_DISCOVERY_THRESHOLD=60         # score to accept candidate (0-100) — lower than Lidarr since movie tastes are broader
+    RADARR_DISCOVERY_THRESHOLD=50         # score to accept candidate (0-100) — lower than Lidarr since diverse seeds rarely overlap
     RADARR_DISCOVERY_LOOKBACK_DAYS=30     # Emby watch history window in days (movies rewatched less often)
     RADARR_DISCOVERY_MAX_SEEDS=5          # max seed movies from Stage 1
     RADARR_DISCOVERY_MAX_ADDS=5           # max movies to add per run