Normalize YouTube Shorts links to watch URLs

docs/media-scraper.md

@@ -96,7 +96,10 @@ One SQLite file per scrape (`index.db`, `node:sqlite`).
 slopbot backend could ask for it (one request; null otherwise). Indexes: `kind`, `posted_at`, and
 `media_sources(message_id)`. The database is an index; the media
 folder holds bytes. Every item carries its poster (`author_id`, `author_name`), the post time
-(`posted_at`), and its dimensions (`width`, `height`, `aspect_ratio`). Embed rows also store
+(`posted_at`), and its dimensions (`width`, `height`, `aspect_ratio`). `youtube.com/shorts/<id>`
+links are stored as `watch?v=<id>` (extraction and a one-time `user_version` migration), because the
+embedded player and View full need the watch form; a shorts row that collides with an already
+indexed watch URL folds into it (preview/file fields merge, message links move over). Embed rows also store
 Discord's `title` (YouTube and friends) plus `title_key`, the lowercased, punctuation-collapsed
 form used by the title search: multi-word prefix tokens by default ("Ma Ca Ju" finds *Man who jumps
 over car*), one phrase with the exact-match toggle, and rows without a title drop out while a