Parent

#15

What to build

Introduce Extractor as a plain struct in src/runner/extract.rs. Move link extraction and asset classification logic out of crawler.rs (and/or wrap the existing src/extract/ module) behind this single seam. Interface: fn extract(&self, response: &FetchResponse) -> ExtractedContent where ExtractedContent carries Vec<DiscoveredUrl> and asset classifications.

No trait — one impl, pure logic. Crawler::process_job calls Extractor directly after a successful fetch.

Acceptance criteria

• Extractor struct in src/runner/extract.rs
• Link extraction + asset classification flow through Extractor
• Crawler::process_job calls Extractor instead of inline helpers
• Inline helpers in crawler.rs removed if fully subsumed
• #[cfg(test)] mod tests with HTML fixtures: relative URLs, base href, srcset, link rels
• NDJSON regression test from runner: bootstrap module shells + NDJSON regression test harness #16 still passes byte-for-byte
• cargo test --all-features green

Blocked by

• runner: bootstrap module shells + NDJSON regression test harness #16