rocksdb-resharding: add a retry/until

When retrieving the container_image, if the corresponding OSD
isn't running (not started/ready yet), let's retry during 120sec before
we let the playbook fail.

This is because of a failure seen in the CI like following:

```
TASK [get container image currently used by osd container] *********************
task path: /home/jenkins-build/build/workspace/cephadm-ansible-prs-el8-functional/rocksdb-resharding.yml:66
Monday 13 June 2022  15:20:38 +0000 (0:00:00.015)       0:00:00.683 ***********
ok: [localhost -> ceph-node0] => changed=false
  cmd:
  - cephadm
  - shell
  - ceph
  - orch
  - ps
  - --daemon_type
  - osd
  - --daemon_id
  - '0'
  - --format
  - json
  delta: '0:00:02.669373'
  end: '2022-06-13 15:20:42.149214'
  rc: 0
  start: '2022-06-13 15:20:39.479841'
  stderr: |-
    Inferring fsid 4217f198-b8b7-11eb-941d-5254004b7a69
    Inferring config /var/lib/ceph/4217f198-b8b7-11eb-941d-5254004b7a69/mon.ceph-node0/config
    Using ceph image with id '7d10b4103611' and tag 'master' created on 2022-05-23 21:52:02 +0000 UTC
    quay.ceph.io/ceph-ci/ceph@sha256:0ece388ce186bf2122eb4f3389d6b108fa94aa3541d40d08449286fffc34e29f
  stderr_lines: <omitted>
  stdout: |2-

    [{"daemon_id": "0", "daemon_name": "osd.0", "daemon_type": "osd", "events": ["2022-06-13T15:20:20.821851Z daemon:osd.0 [INFO] \"Deployed osd.0 on host 'ceph-node4'\""], "hostname": "ceph-node4", "is_active": false, "memory_request": 4294967296, "ports": [], "service_name": "osd.osd", "status": 2, "status_desc": "starting"}]
  stdout_lines: <omitted>
```

We can see the 'status_desc' is 'starting'.

Signed-off-by: Guillaume Abrioux <[email protected]>

rocksdb-resharding.yml

@@ -67,6 +67,9 @@
           command: "{{ cephadm_cmd }} orch ps --daemon_type osd --daemon_id {{ osd_id }} --format json"
           changed_when: false
           register: ceph_orch_ps
+          retries: 120
+          delay: 1
+          until: (ceph_orch_ps.stdout | from_json)[0]['status_desc'] == 'running'
 
         - name: set_fact container_image, container_host
           set_fact: