fix: wait through transient argo degradation

[bxvtr]

Summary

• Fresh Greenfield acceptance showed monitoring transition OutOfSync/Missing → OutOfSync/Degraded → later Synced/Healthy with no operator repair.
• A single valid Degraded snapshot is therefore not reliably terminal during first reconciliation.
• All valid non-final Application states now use the existing bounded wait; only all Synced/Healthy is success.
• Malformed, empty, or structurally invalid Application evaluation still fails immediately.
• Persistent Degraded still fails through the unchanged bounded timeout (CLEAN_ROOM_ARGO_ATTEMPTS=60, CLEAN_ROOM_ARGO_DELAY=15).
• No timeout increase, no desired-state change, no auto-repair, and no monitoring-specific special case.
• PR fix: handle transient clean-room convergence states #71 live Ansible-output behavior remains unchanged.

Test plan

• Required PR CI green on this exact head
• Live branch acceptance of fix/clean-room-argo-degraded-wait before merge
• Confirm deploy waits through transient OutOfSync/Degraded rather than aborting
• Confirm only all Synced/Healthy produces Argo PASS, and persistent non-final state times out at the existing bound