fix(nuq-postgres): reap orphaned NULL times_out_at backlog rows (firecrawl#3413)

nuq_queue_scrape_backlog_reaper previously only deleted rows where
times_out_at < now(), leaving any row with times_out_at = NULL in place
forever. Combined with nuq_group_crawl_finished's NOT EXISTS check on
queue_scrape_backlog, a single orphaned NULL row on a group blocks its
crawl from ever flipping to 'completed'.

Add a fallback clause: also delete NULL-timeout rows whose created_at
is older than 25h (the crawl Redis state TTL + 1h buffer). This unwedges
existing stuck groups and protects against any producer that forgets to
set times_out_at in the future.

Paired with the queue-jobs.ts fix that stops producing NULL times_out_at
for crawl-linked backlog rows.

Co-authored-by: firecrawl-spring[bot] <254786068+firecrawl-spring[bot]@users.noreply.github.com>
Co-authored-by: micahstairs <micah@sideguide.dev>

Author: firecrawl-spring[bot]

apps/nuq-postgres/nuq.sql

@@ -120,8 +120,13 @@ SELECT cron.schedule('nuq_queue_scrape_lock_reaper', '15 seconds', $$
 $$);
 
 SELECT cron.schedule('nuq_queue_scrape_backlog_reaper', '* * * * *', $$
+  -- Also sweep orphaned rows with NULL times_out_at older than the crawl
+  -- Redis TTL window, so a missing timeout on a backlog row can't wedge
+  -- nuq_group_crawl_finished forever via its NOT EXISTS check on backlog.
   DELETE FROM nuq.queue_scrape_backlog
-  WHERE nuq.queue_scrape_backlog.times_out_at < now();
+  WHERE nuq.queue_scrape_backlog.times_out_at < now()
+     OR (nuq.queue_scrape_backlog.times_out_at IS NULL
+         AND nuq.queue_scrape_backlog.created_at < now() - interval '25 hours');
 $$);
 
 SELECT cron.schedule('nuq_queue_scrape_reindex', '0 9 * * *', $$