fix: call replication/failback on target cluster/pool/volume resolved…

[geoffrey1330]

… via relationship lookup

fix: failback correctness — slot flip-flop, DeleteSource, and test coverage

Problem

After a failover, triggering a failback left volumes stuck or oscillating:

1. Wrong cluster targeted — the ops controller called the failback API on the source cluster instead of resolving the relationship to find the target cluster where the failed-over volume actually lives.
2. Slot stuck in FailedOver — reconcileSyncStatus had no handler for cutover_done/failed_over with is_source: false, so the slot never transitioned back to replicating/source.
3. Slot flip-flop — reconcileReplicating unconditionally set cutover_done/target when it saw a done state, racing against reconcileSyncStatus and causing infinite oscillation between the two states.
4. do_replicate not restored — the original source lvol had do_replicate cleared during failover; _swap_failback_lvol_uuid blindly copied that False value, so replication never resumed after failback.

Solution

• Route failback API calls via relationship lookup to the correct cluster/pool/volume.
• Add is_source: false detection in both reconcileSyncStatus and reconcileReplicating — when the backend reports a terminal cutover state with is_source: false, both paths now agree to write replicating/source, stopping the oscillation.
• Drive failback slot state via slot controller polling instead of premature ops controller patches.
• Extend DeleteSource to cover failback: deletes the failed-over volume on the target cluster when set (previously only worked for migration).
• Restore do_replicate = true and replication cadence config from the failback source after UUID swap.
• Extract backendStateFailedOver constant (goconst lint).

Tests

• Fixed TestSlot_ReconcileReplicating_FailedOver_UpdatesState and TestSlot_ReconcileReplicating_DetectsCutoverDone_SetsDirectionTarget — both had IsSource: false but were testing the forward failover path; corrected to IsSource: true.
• Added TestSlot_ReconcileReplicating_CutoverDone_Failback_SetsReplicatingSource and TestSlot_ReconcileReplicating_FailedOver_Failback_SetsReplicatingSource covering the is_source: false → replicating/source path.
• Updated ops controller tests to mock relationship lookup.

Test plan

• go test ./internal/controller/... green
• deleteSource: true removes the failed-over target volume (targetLvolID changes post-failback)
• No flip-flop in kubectl get replicationslot -w