Fix stale YouTube title and description after SPA navigation

[vicky469]

After navigating from video A to video B in the same YouTube tab, clipping can combine B's content with A's title and description. YouTube can retain A's Open Graph tags and structured data; when the extractor rejects those values, the generic metadata fallback restores them.

This adds a fallback to the rendered watch title and description, scoped to a ytd-watch-flexy whose video-id matches the requested URL. It uses the expanded description when available, otherwise the rendered snippet. An explicit extractor opt-in prevents empty title/description values from falling back to rejected metadata; other extractors keep their existing behavior.

Follow-up to #174 and #198. Related: obsidianmd/obsidian-clipper#735. Clipper's bug-report guidance directs incorrect-content issues to Defuddle.

Validation:

• Reproduced stale head metadata after real same-tab YouTube navigation. Verified the built parser against the captured rendered DOM with the current JSON-LD omitted to exercise the missing-schema fallback.
• Added an anonymized HTML fixture and regression tests covering sync/async parsing, collapsed and absent descriptions, mismatched video IDs, partial current metadata, and preservation of another extractor's fallback. Observed the regression tests fail before the fix.
• npm test: 585 passed.
• npm run test:jsdom: 585 passed.
• npm run build, npm run size, and npm run lint:innerhtml: passed.