Parent

#15

What to build

Move per-attempt lifecycle event and hook emission from Crawler into JobRunner. Run-level / decision events stay on the Crawler. Then delete dead helpers in crawler.rs that are no longer reachable.

Fires from JobRunner (inside run()): job_started, fetch_first_byte, render_navigated, challenge_detected, fetch_completed, extract_completed, job_failed.

Stays on Crawler: job_enqueued, job_dequeued, job_admitted, job_rejected, job_retried, job_dropped, job_persisted, frontier_seeded, run_started, run_stopped.

NDJSON event names and payloads must remain byte-for-byte unchanged; only the emitter changes. The regression test from #16 is the trip wire.

After events move, sweep src/crawler.rs for helpers that no longer have callers and delete them.

Acceptance criteria

• Per-attempt events listed above fire from inside JobRunner::run via the injected Arc<EventSink> / Arc<HookRegistry>
• Run-level events listed above continue to fire from Crawler
• NDJSON byte-for-byte parity confirmed by the regression test from runner: bootstrap module shells + NDJSON regression test harness #16
• Hook firing points (Rust and Lua) unchanged in event names and HookContext payload
• Dead helpers in crawler.rs deleted (track LOC delta in PR description)
• cargo test --all-features green
• cargo clippy --all-targets --all-features -- -D warnings green
• PR description includes before/after LOC count for src/crawler.rs

Blocked by

• runner: JobRunner::run becomes per-Job entry point in Crawler::process_job #22