feat: add Xiaohongshu (RedNote) extractor

[Afterimages]

Extract note content (description, images, video, tags) and comments from xiaohongshu.com note pages via INITIAL_STATE JSON parsing.

Comment extraction uses a three-tier fallback strategy:

1. Parse from INITIAL_STATE (HTML source / MAIN world bridge)
2. Fetch from XHS comment API (async fallback)
3. Extract from rendered DOM elements (works without MAIN world access)

Additional features:

• Normalize JS-only tokens (undefined/NaN/Infinity) in JSON-like state
• Remove duplicate hashtag mentions from description text
• Include comments in contentHtml by default; opt-out via excludeCommentsFromContent option
• Provide comments as {{comments}} template variable (markdown)
• Register extractor for xiaohongshu.com URL patterns

Relates #273

[kepano]

Please add tests/fixtures.

[Afterimages]

Please add tests/fixtures.

Force-pushed the branch after squashing the changes into a single commit.

Updates since the previous review:

• Added Xiaohongshu fixtures and unit tests
• Improved SPA stale-state handling and DOM fallbacks
• Fixed DOM comment author extraction
• Reduced the bundle-size impact
• Verified the extractor manually in the browser extension against a live Xiaohongshu note

Manual verification confirmed that:

• The title, author, publication date, description, tags, and images are extracted correctly
• Comments and nested replies are preserved
• The extracted content can be saved to Obsidian correctly
• SPA navigation fallback works as expected

All CI checks are now passing on Node.js 20, 22, and 24.

Ready for re-review. Thank you!