blueprint-execution: Break dependency between `PUT /omicron-zones` success and subsequent cleanup steps

[jgallagher]

blueprint-execution has two cleanup steps that depend on the success of the earlier PUT /omicron-zones step:

• zone cleanup
• saga reassignment
• failed support bundle cleanup

All of these steps assume that the zones they are cleaning up after are no longer running, but that's only true today because PUT /omicron-zones is synchronous (i.e., sled-agent only returns success if it has already stopped any zones that shouldn't be running) and because execution is stopped if the PUT /omicron-zones step fails. We definitely want to change the second of those (this is #6999), and longer term we probably want to change the first one too (converting sled-agent into more of a "accept and return the new config, then make it real via a reconciler loop in the background).

#7524 is a small PR that makes these dependencies explicit in code. We should break this dependency somehow:

• The planner could confirm that a zone is gone and indicate it's ready for cleanup via some property in the blueprint (similar to the treatment disks got in Expunge and Decommission disks in planner #7286)
• Is it possible for the executor to know this on its own? (I don't think so but maybe?)

This is a blocker for fixing #6999.

[jgallagher]

Fixed by #7713