feat: prioritize high-value comments for review

[dyxg]

Summary

• Add a configurable priority comment selector that ranks high-engagement posts and comments while preserving the original comment table.
• Route the selected priority queue through LangGraph output, SSE comment streaming, per-comment AI generation, and /insights review data.
• Update the frontend review queue to refresh existing task insights, display selection metadata, and copy generated copy instead of sending comments.
• Add task 3 documentation, Auto-Log records, and focused backend/frontend tests.

MediaCrawler boundary

• Existing /api/ad-intel/run MediaCrawler capability is preserved.
• Automated tests and local mock task validation do not invoke MediaCrawler or require Xiaohongshu login.

Testing

• python -m pytest -q → 63 passed
• npm test → 7 passed
• npm run build → passed
• python -m ruff check app/services/comment_selector.py tests/test_comment_selector.py app/api/routes.py → passed

Notes

• Local ignored mock data was created under data/ for manual testing with mock-priority-001, but it is intentionally not included in the PR.
• Existing dirty vendor/MediaCrawler worktree state was left untouched and is not part of this PR.

[dyxg]

Update pushed in 70547fd:

• GET /api/ad-intel/task/{task_id}/insights review items now include:
  • likes: integer comment like count
  • note_id: source post id string
• Existing comment_like_count remains for backward compatibility.
• Frontend ReviewItem type updated and UI now prefers likes for display.

Validation:

• python -m pytest -q → 63 passed
• npm test → 7 passed
• npm run build → passed
• python -m ruff check app/api/routes.py tests/test_routes_helpers.py → passed