The forge pages the phone once for errors, internal faults, a long queue or a full disk that has held five minutes, without flapping #2492

Second part of #2285, after the drain page lands.

Done looks like:

• [ ] The incident pager owns an explicit list of vitals causes: errors, internal, queue:, disk. Every other cause (drill, mirror, stuck:, red:, snapshots, budget:) is ignored, because something else already pages it or it is a setting, not an incident.
• [ ] An owned cause pages once it has held 5 minutes. It counts as cleared only after it has been absent for 5 minutes (track absentSince per key), and only then sends Recovered.
• [ ] The boot baseline is seeded from the owned causes present at boot whatever their age, so a deploy does not re-page disk.
• [ ] Tests with a fake clock: errors present, absent 1 minute, present again sends one page and no Recovered; absent 6 minutes sends one Recovered; a drill or stuck:r1 cause past five minutes sends nothing; a restart with disk present pages nothing.