[sled-storage] Remove "self-managed" physical disks #5328
Description
As a part of #5172 , we started using a "ledger of control-plane disks" stored on M.2s. This tells sled agents, durably, "which U.2s they should use", as instructed by Nexus. This is stored in a ledger for bootstrapping purposes -- this info needs to be accessible before Nexus/CRDB are up and running.
However, as a bridge to keep "ledger-less" deployments working, a kludge exists in this PR to automatically "self-manage" U.2s on devices which have no ledgers, but do have U.2s with zpools.
This works as a transitional measure, but should be removed once we're confident we've migrated to a world where deployments are using a ledger.