Block agent bootstrap on existing host deployments

[Copilot]

unbounded-agent preflight and start did not detect existing or partially removed host deployments. This could allow bootstrap to continue with stale nspawn or agent state instead of requiring an explicit reset.

• Existing deployment detection

  • Adds a host preflight check for prior deployment artifacts.
  • Detects registered kube1 / kube2 nspawn machines, machine rootfs directories, nspawn configs, service overrides, bpffs paths, applied configs, and agent daemon artifacts.

• Start-time guard

  • Runs the same detection before start performs any host mutation.
  • Returns reset guidance when stale state is found.

• Operator guidance

  • Error output now says a node reset is needed before retrying preflight or start, without hard-coding a product-specific reset command.

Example failure:

existing node deployment detected; node reset is needed before running start again

[bcho]

@copilot please remove hard coded unbounded-ageant reset mentioning

[Copilot]

@copilot please remove hard coded unbounded-ageant reset mentioning

Addressed in f99a6ab by replacing the hard-coded unbounded-agent reset guidance with product-neutral node reset wording.

[bcho]

@copilot test is failing

Error: existing-deployment]: existing node deployment artifact detected (bpffs mount path): bpffs mount path /run/bpffs/kube1; node reset is needed before running preflight or start again (target: /run/bpffs/kube1)

[Copilot]

@copilot test is failing

Error: existing-deployment]: existing node deployment artifact detected (bpffs mount path): bpffs mount path /run/bpffs/kube1; node reset is needed before running preflight...

Fixed in 7f1cf3f by adding bpffs mount cleanup to daemon AgentReset, so reset removes /run/bpffs/kube1/kube2 before e2e re-runs preflight.