TRR BACKEND: Validate cast/image views for duplicates + performance

[therealityreport]

Goal: Confirm whether core.v_episode_cast, core.v_season_cast, and core.v_person_images produce duplicates or need performance tuning.

Note: Only pursue if duplicates are actually observed.

TODO

• Run data reality check SQL snippets:

  -- Episode cast duplicates by (episode_id, person_id, credit_category)
  SELECT episode_id, person_id, credit_category, COUNT(*)
  FROM core.v_episode_cast
  GROUP BY episode_id, person_id, credit_category
  HAVING COUNT(*) > 1;
  
  -- Season cast duplicates by (season_id, person_id)
  SELECT season_id, person_id, COUNT(*)
  FROM core.v_season_cast
  GROUP BY season_id, person_id
  HAVING COUNT(*) > 1;
  
  -- Person images duplicates by (person_id, media_asset_id, kind)
  SELECT person_id, media_asset_id, kind, COUNT(*)
  FROM core.v_person_images
  GROUP BY person_id, media_asset_id, kind
  HAVING COUNT(*) > 1;

• If duplicates exist and are problematic:
  • update views with SELECT DISTINCT and document why
  • or add constraints/indexes where appropriate
• If duplicates exist but are harmless:
  • document that consumer must dedupe and close the issue
• Add lightweight EXPLAIN plans + indexes if queries are slow

Acceptance

• Views are correct for Screenalytics candidate selection and perform acceptably on production data
• Duplicate behavior is documented (either fixed or consumer-responsibility)