In-use volume does not support the same ceph cluster retype

[wang]

Description：
  Ceph storage supports different types of disks, including SATA, SAS, and SSD, which are divided into SATA SSD and NVME.In the production, disks with different media are divided into different pools. In cinder, there should be different volume types. The corresponding service products are ordinary cloud disk, efficient cloud disk and SSD disks.Switch normal cloud disk to efficient cloud disk according to business requirements, so that data can be migrated in real time between different pools without data loss.
 The current Openstack environment does not support the retype (cross-pool) operation of the volume of in-use state in the ceph cluster. Libvrit does not provide a data migration interface. When rebase, the storage type of ceph is not supported, and the cinder-api does not support the retype operation of the volume of in-use state.

Environment：
1. openstack-Queens
2. hypervisor: Libvirt + KVM
3. storage type : Ceph
4.Cinder concatenates different pools
  Cinder creates the volume type of the corresponding pool

  cinder extra-specs-list
+--------------------------------------+-----------+--------------------------------------+
| ID | Name | extra_specs |
+--------------------------------------+-----------+--------------------------------------+
| 008d4021-8010-43b5-8667-ba028ee74fe9 | ceph-test | {'volume_backend_name': 'ceph-test'} |
| 50a3919c-e722-4a1e-aa42-c8b1303320f9 | ceph | {'volume_backend_name': 'ceph'} |
+--------------------------------------+-----------+--------------------------------------+

Do the below test：
1.Create instance and volume
2.Attach the volume to the server
3.Execute retype: cinder retype test-retype ceph-test --migration-policy on-demand

Actual result：
Pool hasn't changed