Etcd doesn't have persistent data storage

Currently 'etcd' service is spawned as replica set without persistent volumes and only 1 pod is scheduled simultaneously. So if the pod with etcd is deleted, then no OpenStack service can start:

> Dependency "xyz" is not ready yet

Steps to reproduce:

1. Deploy k8s with ccp-installer
2. Download fuel-ccp
3. Build images by fuel-ccp
4. Deploy OpenStack in 'ccp' namespace of k8s
5. Delete 'etcd' pod in 'ccp' namespace and wait until it's re-created
6. Restart 'keystone' service (delete pod)

Expected result: keystone is restarted and works fine

Actual result: keystone isn't started because 'start_script.py' is infinitely waiting for dependency "keystone-db-create" (the key doesn't exist in etcd)