fix: improve content root and lead extraction

[ryanfowler]

Summary

• avoid bibliography/reference sections replacing article content
• preserve lead media without injecting page titles into Markdown content
• remove explicit terminal related sections and preserve transparent list wrappers
• keep resolved page titles in metadata, consistent with Readability-style content output
• update regression snapshots and tests

Validation

• cargo fmt --check
• cargo clippy --all-targets --all-features -- -D warnings
• cargo test

The quality runner still reports the pre-existing app-empty-shell expected-failure gate; the same failure reproduces at the base commit.