Improve documentation content extraction

[ryanfowler]

Summary

• Remove redundant disclosure labels from API documentation while keeping authored details labels.
• Remove terminal feedback and page-maintenance panels with structural evidence. Keep authored footer text.
• Extract combined publication dates and authors from compact article bylines. Reject hidden and sidebar metadata.
• Use precomputed metadata visibility, context, and positions to keep the new fallback linear.
• Add positive and negative regression coverage for documentation cleanup and metadata inference.

Research

I sampled current article, documentation, repository, issue, and pull request pages. The sample included links from Lobsters, rustdoc, MDN, Kubernetes, Rust project pages, Python and Go documentation, and GitHub. The changes target patterns that occurred across page generators. They do not add site-specific selectors.

Validation

• cargo fmt --check
• cargo fmt --manifest-path cli/Cargo.toml --check
• cargo clippy --all-targets --all-features -- -D warnings
• cargo clippy --manifest-path cli/Cargo.toml -- -D warnings
• cargo test
• node tools/extractor-eval/index.mjs --all --gate legible (134 fixtures)
• cargo bench --bench smoke

Three independent review passes completed. All reported findings were addressed before this pull request.