v5.8: Report PROCESSING jobs running longer than 24 hours in the monitor

The reaper has one blind spot: a callback stuck in an infinite loop
that keeps heartbeating through the middleware never ages in updated_at,
so stalledJobTimeout never fires and the job sits in PROCESSING forever
- silently blocking its serialGroup. The monitor report now includes a
third category: PROCESSING jobs whose run (last_attempt_at) started
more than 24 hours ago. Their presence, like PERMANENTLY_FAILED, raises
the report level to critical since neither fixes itself.

reportFailedJobs() is renamed to reportStuckJobs() to match what it now
reports (the method was introduced within this release, so no BC
concern), and the per-state 'oldestCreatedAt' key becomes 'oldestSince'
- for PROCESSING it carries the start of the run, not the job creation.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>