[reconfigurator] Stop reading dataset UUIDs from datastore

[smklein]

In #6229 , nexus begins managing datasets via the reconfigurator. As a part of this logic, when ensuring a dataset exists, the reconfigurator first checks for the dataset within the datastore, to ensure that if it already exists, the UUID is re-used.

As an example:

• Suppose we have a deployed disk with a Crucible UUID of xyz on a physical disk
• Suppose we update the deployment to get this PR
• Suppose the reconfigurator runs, sees the physical disk, and sees that it needs a Crucible dataset + zone.

If the reconfigurator allocated a new UUID for Crucible, we'd end up with two Crucible datasets on the sled - this would be wrong. Instead, it looks for any "Crucible kind" datasets on the zpool within that disk, and re-uses the existing UUID.

If the reconfigurator was confident that "all existing datasets existed in the blueprint", this logic would be unnecessary -- instead of reading from the datastore, the reconfigurator could look for old UUIDs by scanning the parent blueprint.