Implement crawl resume against the Crawl registry (ADR-0007)

[filipeforattini]

What to build

Implement crawl resume against the Crawl registry: look up the Crawl, reopen its Frontier and Seen-set, and continue draining pending Jobs as a new run over the same crawl_id (replacing the current stub). Detect a stale-running Crawl (process died) and take it over rather than double-run. See ADR-0007.

Acceptance criteria

• crawl resume reopens a registered Crawl and drains its pending Jobs.
• Resume starts a new run_id over the same crawl_id.
• A stale-running Crawl is taken over, not double-run.
• Tests cover resume of an interrupted Crawl.

Blocked by

• Crawl registry + crawl list (ADR-0007) #65