feat(convert): add structured metadata extraction from HTML pages

[chaliy]

What

Add PageMetadata struct and extraction from HTML pages during fetch. Returns structured metadata alongside converted content in FetchResponse.metadata.

Why

Agents currently have to re-parse markdown to get basic page info like title, description, and links. This is the single biggest improvement for agentic use — agents need this metadata universally.

How

• New PageMetadata struct with: title, description, language, canonical_url, author, published_date, modified_date, links (Vec), headings outline
• extract_metadata() — single-pass HTML parser for meta tags, title, links, language, canonical URL
• extract_headings() — separate pass for heading outline extraction
• Both integrated into DefaultFetcher — metadata populated when HTML content detected
• OG tags override basic HTML tags (og:title > title, og:description > meta description)
• DoS limits: max 500 links, max 200 headings per page

Risk

• Low — additive change, existing behavior unchanged
• New optional field on FetchResponse, backward-compatible

Checklist

• Unit tests passed (16 tests covering all metadata fields)
• Clippy clean
• Docs build clean
• Specs are up to date

Closes #71