feat: add agent content focus extraction

[chaliy]

What

Add low-noise content_focus modes for AI agents: readable and agent. The default fetcher now reports the extraction method in page metadata and MCP/help/docs expose the option. Also relaxes the git attribution guidance so GIT_USER_NAME/GIT_USER_EMAIL are not a blocker and YOLOP attribution is allowed when requested.

Why

FetchKit is primarily an agent tool, so callers need cleaner content than full-page HTML conversion when pages include nav/sidebar/footer noise. Contributors also should not be blocked by env-only git user setup when normal git config is present.

How

Implemented deterministic readability scoring over bounded semantic/content containers, with fallback to existing semantic boilerplate stripping. Added unit and fetcher coverage for the new extraction path, updated specs, README, and tool help.

Risk

• Low / Medium
• New behavior is opt-in via content_focus; default full behavior remains unchanged. Heuristic extraction may miss some pages and fall back to main.

Checklist

• Unit tests are passed
• Smoke tests are passed
• Documentation is updated
• Specs are up to date and not in conflict