feat(core): offline rebuild for unpublished degraded volumes

[yugchaudhari]

Iteration 2 of offline volume rebuild (openebs/openebs#4208), builds on the detection-only reconciler from #1103.

When an unpublished volume goes degraded, the reconciler now creates a temporary unshared nexus (target_config with protocol: None) after a grace period. That makes the volume look published to the existing HotSpareReconciler, which rebuilds the faulted replicas. Once the volume is Online again, the reconciler tears the nexus down and the volume returns to unpublished.

Key point: no rebuild logic is reimplemented, we lean on HotSpare. We just stand up and tear down the nexus.

New config: --offline-rebuild-grace-period (default 10m), behind the existing --offline-rebuild-enabled flag.

BDD tests cover happy path, feature-disabled, and never-published precondition.

[yugchaudhari]

Tested this on a 3-node cluster, both the BDD suite and a manual run.

BDD tests pass (cargo test -p agents --test core offline_rebuild): happy path, feature-disabled, and the never-published precondition.

For the manual run I created a 2-replica volume, published then unpublished it (to establish the health info), and killed one of the replica's io-engine nodes. The volume state over time:

Degraded  target=None                ← node killed, volume degraded
Degraded  target=io-engine-3/none    ← offline rebuild created the temp unshared nexus (protocol=none)
Online    target=None                ← rebuild done, nexus torn down, back to unpublished
Online    target=None                ← stays healed

And the reconciler's own logs line up with that:

16:39:39  DEBUG  Offline rebuild waiting for grace period, remaining: 7.97s
16:39:46  DEBUG  Offline rebuild waiting for grace period, remaining: 956ms
16:39:47  INFO   Initiating offline rebuild: creating non-shared nexus
16:39:47  INFO   Offline rebuild nexus created; HotSpareReconciler will handle the rebuild
16:39:50  INFO   Offline rebuild complete; tearing down temporary nexus
16:39:50  INFO   Temporary nexus destroyed; volume returned to unpublished state

So the full lifecycle works end to end: degraded unpublished volume → temp nexus → HotSpare rebuild → teardown → back to Online/unpublished.