need a built-in "factory reset" (clean slate)

Oxide currently has at least four different "clean slate" scripts:

• one used by support
• one in rkadm
• one in a branch of rkdeploy that's currently deployed and used for production systems
• one in meta

These are all inconsistent. This speaks to the need for having a single implementation that can be shared. We've long talked about having something like this in Wicket.

Relatedly: 3 of these have a bug in which they do not wipe the update dataset, which absolutely needs to be wiped. This has resulted in a significant customer issue. This suggests another thing that's been mentioned before, which is that it would be better if the clean slate script had an allow-list of things not to wipe rather than only wiping specific, known directories and files. (Even better would be to reconstruct the pool from first principles, but I don't know if that's feasible.)

Compare the clean-slate.sh implementations in customer-support and rkadm with the deployed rkdeploy version and the meta/engineering/rack-support/clean-slate.adoc guidance. Determine the Wicket entry point and the required wipe policy, including the update dataset. Done means a single shared factory-reset implementation replaces the inconsistent scripts and reliably preserves only explicitly allowed data.