kv,storage: ability to decommission stores

Currently, we support decommissioning one or more nodes, e.g. `cockroach node decommission N1 N2 ...`.

When a node has multiple stores attached, this will effectively decommission _all_ of the stores attached to it.

Improve the tooling to support decommissioning by _store_. This would allow, for example, removing of a faulty block device from a node without needing to decommission all other stores on the node.

Given the existing tooling, we'd probably want to think about the UX. Maybe something like a `cockroach decommission --store S1 --store S2 ...`?

**Note also the possible interaction with WAL failover. Before a store is removed, its "partner store" must stop relying on it for WAL failover.**

Jira issue: CRDB-42552

Epic CRDB-48087