fix(replica-scheduling): power off replica node should not rebuild new replica on same node (backport #2918)

[mergify[bot]]

Which issue(s) this PR fixes:

Issue longhorn/longhorn#1992

What this PR does / why we need it:

• Before creating a new Replica CR, verify the availability of suitable nodes with sufficient disk to accommodate the replica. This prevents creating Replicas that cannot be scheduled.
• During node disk candidate selection, exclude nodes where the InstanceManager is not in a running state. This prevents creating redundant Replica CRs on nodes that haven't fully recovered after a power down.

Special notes for your reviewer:

None

Additional documentation or context

None

---

This is an automatic backport of pull request #2918 done by [Mergify](https://mergify.com).

[c3y1huang]

Ref https://github.com/longhorn/longhorn/issues/8859

[c3y1huang]

TODO:

• [ ] Backport https://github.com/longhorn/longhorn-manager/pull/2939
• [ ] Backport https://github.com/longhorn/longhorn-manager/pull/2941
• [ ] Backport https://github.com/longhorn/longhorn-manager/pull/2942
• [ ] Backport https://github.com/longhorn/longhorn-manager/pull/2950