v0.11.0

Stop waking a pool for a run that will never start

A GitHub run can sit queued for ever with no jobs attached. Autoscale
counted it, so the pool woke every twenty minutes, found nothing, and
stood down again, for as long as the run existed. Nothing reported it,
because a pool that wakes and stands down looks healthy.

Such runs are now subtracted from the queued count rather than
suppressing the pool, so every other queued run still wakes it. doctor
names the run and how to cancel it, status counts them, and the notifier
reports each one once. Tune with RUNPOOL_STUCK_WAKES, or set it to 0 to
disable the guard.

No action needed on upgrade.