I'm guessing that disabling the archiver after having it enabled is not
a common case, but with this we should be able to remember how to clear
out the alert.

I'm not 100% certain about what to do in the case that the archive is
valid, so I indicated that maybe we want to extract it in a subdir. we
can adjust that when we find a case where that happens. today's case was
a corrupted archive that we could delete.

this is in addition to what we can find with `gnt-instance info`

That'll make it easier to identify what are the obsolete packages in the
output. The openmetrics output is not needed during investigation.

This should be slightly more readable

credit to lavamind for this procedure, see team#41396#note_2966380

See: team#41969

Closes: tpo/tpa/puppet-hiera-enc#1

cf. team#41991

We stop short of rewriting all playbooks for Prometheus, and instead
add references to the task of adding playbooks for
everything (prometheus-alerts#16) where we found references to
nagios.

I forgot to change it after the copy/paste