Harden rollout VM lab startup

[chrisbliss18]

Summary

This hardens the rollout VM lab so snapshot-backed rehearsal targets can start from an offline lab topology without hanging in the staging step.

The lab harness now adds start-topology and calls it from install-v2, so the db/v1/v2 lab guests are started before SSH waits and artifact staging begin. The helper is intentionally narrow: it only touches the three prefix-derived lab domains, validates that the complete db/v1/v2 topology exists before starting anything, auto-starts only cleanly shut off VMs, and refuses unexpected states such as crashed, paused, or suspended for operator inspection.

The VM lab docs were updated to make that behavior explicit near the setup instructions and snapshot-backed flow notes.

Why

During rollout readiness rehearsal, make rollout-vm-lab-snapshot-all-smoke assumed the VMs were already running before staging v2. When the baseline snapshots left domains shut off, the target could sit quietly waiting for SSH. This change makes the rehearsal target self-contained while keeping VM lifecycle automation tightly scoped and easy to audit.

Example output

Powered-off lab topology now starts explicitly before staging:

INFO start_topology prefix=jetmon-rollout vms=jetmon-rollout-db,jetmon-rollout-v1,jetmon-rollout-v2
PASS vm_started=jetmon-rollout-db previous_state="shut off"
PASS vm_started=jetmon-rollout-v1 previous_state="shut off"
PASS vm_started=jetmon-rollout-v2 previous_state="shut off"

Wrong prefix or missing topology fails before any start attempts:

INFO start_topology prefix=jetmon-missing vms=jetmon-missing-db,jetmon-missing-v1,jetmon-missing-v2
WARN missing_vm_domain=jetmon-missing-db
WARN missing_vm_domain=jetmon-missing-v1
WARN missing_vm_domain=jetmon-missing-v2
FAIL topology is incomplete; run create-topology first or check JETMON_ROLLOUT_PREFIX=jetmon-missing

Unsafe libvirt states are refused for inspection:

FAIL VM is not in a safe auto-start state: jetmon-rollout-v2 state="paused"

Validation

• bash -n scripts/rollout-vm-lab.sh
• git diff --check
• make rollout-vm-lab-stage-v2 from a powered-off topology
• bad-prefix start-topology simulation
• paused-VM start-topology simulation
• make rollout-vm-lab-snapshot-execute-smoke
• make rollout-docs-verify