fix icebox migration

The second job here would have never ran: the first job would have
repeatedly warned about stale tickets, constantly reseting the timer.

By using the (newly created) ~Stale label, we ensure that

1. the first job doesn't constantly reset the timer (which would have
   been really annoying)

2. that the second job can actually run

This implies reducing the interval on the second job, because it
starts counting after the first one has run.

This also implies that, to reset the *first* timer, the `Stale` label
needs to be removed, which the bot now does as well, if a ~"Needs
Information" ticket gets updated.

(Note that we may want to generalize this to other labels or all
labels, but for now we're keeping the scope limited to ~"Needs
1 job for tpa-triage in 49 seconds (queued for 28 seconds)