docs(roadmap): track #309 and #311, and record #292 as done

[Gerrrt]

Adds #309 and
#311 under Monitoring, and
moves #292 into Done.

Why #292 moves at the same time

It shipped — #302 and
#306 — and its entry was still
sitting in Monitoring describing the work as pending. Leaving it there while
adding its two follow-ups directly above would have been the failure the #97
entry says this section exists to prevent: "The entry outlived the work by a
day."

The Done entry records the part no plan predicted. Deploying the check found
that sysDescr is not morpheus-only — the ilo module walks it too, so the
query returns two series with shiva first — and the runbook comparison was
selecting by parse order rather than by device. It was correct only because the
iLO firmware is 2.82, two components rather than three. An iLO numbered
2.82.1 would have reported the BMC's firmware under the firewall's name.

The two that replace it

• Check shiva's iLO firmware version against the docs too #309, shiva — parsing. It already has both halves of a checkable claim:
  network.md gives it iLO 2.82 and its sysDescr is already scraped. The
  only thing keeping it unchecked is a scoping line, but it is not a one-line
  change, because os_key misreads both sides and the iLO's 4 is a
  generation, not a firmware version.
• Collect the Proxmox VE version from Saruman so the docs can be checked #311, Saruman — collection. Nothing on the wire carries the number the
  documents claim, so Proxmox VE 9 is currently unfalsifiable. The textfile
  collector is already enabled and provisioned; the awkward half is that the
  monitoring host cannot reach VLAN 30, so whatever writes the metric has to
  live on a host this repository can only reach through a person.

Both entries say why they are not trivial, since both look trivial from the
outside — which is the specific way each would get picked up and abandoned.

Docs only. make lint and make check-docs pass.

🤖 Generated with Claude Code