[Feature] 😄 Improve the logic of rolling update

[zhl003]

When the cluster is rolling update:

1. Operator check cluster's GTID are all the same, if not, prevent doing rolling update and rewrite the failed reason to CR status
2. if OK, doing the update
3. Remove useless backup logic of init-sidecar
   mysqlcluster,sidecar: support Clone initial when add new pod, and fix bug after extend pvc, restore from backup fail #250, #370 #291