feat(item-detail): section-aware rendering on ItemDetail page

Structured sections (item_metadata._sections from Phase 1) now drive
the item detail view. ItemDetail picks a renderer based on the
section shape it sees:

- Forum view (Reddit, HN, Mastodon, Threads): post card + indented,
  collapsible comment tree with score + author + depth-based indent.
  Quoted tweets render with an indigo accent border.
- Q&A view (Stack Overflow, Stack Exchange): question card + accepted-
  answer highlighted in green with a check badge + remaining answers
  stacked score-sorted + inline comments.
- Video view (YouTube, TikTok): description block, then chapters with
  timestamp badges, each showing its grouped transcript segments with
  left-rail mm:ss timestamps.
- Paged view (PDF): one page block per page_no with page anchor; body
  sections render their heading/paragraph/code/list structure inside.
- Article view (Medium, Substack, LinkedIn Pulse, generic web): flat
  reading layout with heading levels, paragraphs, block quotes, code
  fences, OCR/alt-text callouts, and an amber "limited extraction"
  pill when metadata_block sections appear.

Legacy items without _sections keep rendering via ContentRenderer —
the new SectionRenderer only activates when structured data is
available, so nothing regresses for existing content.

Frontend build + typecheck clean. 256 backend tests still green.

Author: prakersh