fix(vm): keep target PVC when migrating VM with local disks

[loktev-d]

Description

Two fixes making live migration / Evict of VMs with hotplugged local-storage disks reliable. Follow-up to #2508.

1. vm controller: stable volume swap. The migration volume/affinity were gated on the VD Migrating condition, which is False (WaitForTargetReady) until the target PVC binds. Mid-migration the controller read that as "not migrating" and reverted the hotplug volume/affinity to the source PVC. Now gated on the stable MigrationState.
2. vd controller: no stale target PVC. Back-to-back migrations could adopt a previous migration's still-terminating PVC as the target. Now terminating non-source PVCs are filtered out, so a fresh target is created.

Why do we need it, and what problem does it solve?

VMs with local-storage hotplug disks couldn't be reliably evicted/migrated.

What is the expected result?

Repeatedly Evict a VM with a hotplug disk on a local WFC storage class - target launcher and attachment pod schedule on the target node and the migration completes; no Unschedulable, no WaitForTargetReady stall.

Checklist

• The code is covered by unit tests.
• e2e tests passed.
• Documentation updated according to the changes.
• Changes were tested in the Kubernetes cluster manually.

Changelog entries

section: vm
type: fix
summary: keep target PVC when migrating VM with local disks