feat(clone): add cloneFullCopy parameter for snapshot-based volume cloning

[kvaps]

Summary

• Add a new StorageClass parameter linstor.csi.linbit.com/cloneFullCopy
  that changes Clone() to use snapshot + restore instead of LINSTOR rd clone
• When enabled, clones are created as fully independent copies placed on
  different nodes from the source, avoiding thin pool exhaustion on source nodes
• Migration is best-effort: if no alternative nodes exist, the clone stays
  co-located but remains a full independent copy

Motivation

LINSTOR's built-in rd clone creates thin COW snapshots pinned to the same
nodes as the source volume. This works well for occasional clones, but becomes
problematic when cloning golden images (e.g. VM disk templates) at scale — all
clones accumulate on the source nodes, eventually exhausting the thin pool.

With cloneFullCopy: "true", the CSI driver instead:

1. Creates a temporary LINSTOR snapshot from the source
2. Restores it via VolFromSnap (which uses autoplace for new placement)
3. Migrates clone data away from source nodes (Autoplace with NotPlaceWithRsc)
4. Waits for DRBD sync to complete on new nodes
5. Removes the source-node copies

This produces fully independent volumes distributed across the cluster.

Usage

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: linstor-distributed
provisioner: linstor.csi.linbit.com
parameters:
  linstor.csi.linbit.com/cloneFullCopy: "true"
  linstor.csi.linbit.com/storagePool: "pool"

Test plan

• Clone a volume with cloneFullCopy: "true" — verify clone lands on different nodes
• Clone without the parameter — verify original rd clone behavior is unchanged
• Clone with cloneFullCopy: "true" on a fully saturated cluster (replicas = node count)
  — verify graceful fallback (clone co-located but functional)
• Restart CSI driver mid-clone — verify leftover snapshot cleanup on retry