Implement new mechanism for populating `rack` table to support multi-rack

We currently depend on PopulateRack in Nexus to insert a row into the rack table, and Nexus does this using its own local rack ID. In multi-rack, this will not be wrong per se, but it won't be sufficient since not every rack will have a Nexus instance running on it. We'll need to create some other mechanism to handle the insertion of rows into the rack table without depending on the locally configured rack ID, and we'll have to figure out how to trigger this mechanism when racks that don't have any Nexus nodes running on them join a cluster.

Not sure what the best way to implement this is, and whether it will still use the Populate trait or some other framework, but I'm creating this issue as a placeholder for whatever work will be required to design and implement this functionality.

Start by tracing Nexus's PopulateRack mechanism and the rack table insertion path, then examine whether the Populate trait can support racks without Nexus nodes. Define how cluster events trigger insertion for those racks; done means a designed and implemented mechanism populates every joining rack without relying on a locally configured rack ID.