Fix video ID extraction for reordered YouTube watch URLs

[R4ULtv]

Motivation

• The previous scraper used a strict regex that only matched /watch?v=<id> when v was the first query parameter, which caused valid video links like /watch?pp=...&v=<id> to be skipped.
• Make ID extraction more robust to different YouTube URL shapes so repeated videos are not missed.

Description

• Add an extractVideoId(href: string) helper in src/scraper.ts that parses the URL with new URL(...) and uses searchParams.get("v") with an 11-character validation.
• The helper also supports alternative path-based forms by checking for /watch/<id> in the pathname.
• Update scrapeVideoIds() to use extractVideoId() instead of the previous regex, keeping existing playlist filtering and deduplication logic unchanged (file modified: src/scraper.ts).

Testing

• Ran bun run check (runs tsc --noEmit, oxlint, and oxfmt --check) and all checks passed successfully.

---

Codex Task