[feature] Support change replicas without restart cluster.

[runkecheng]

Is your feature request related to a problem? Please describe.

Now use the postStart script to add nodes to the xenon cluster. This method will cause the cluster to be restarted when the number of replicas is changed.

Describe the solution you'd like

Add relevant logic to the operator instead of script.

step:

1. Set the cluster status to Scale In / Scale out when Replicas changes.

2. Wait for all Pods ready to make RAFT's master-slave status (Previous practice is starting at the Xenon container of each Pod)

3. If all Pod are new (init or closed), set the Pod of the serial number 0 to the leader directly

Describe alternatives you've considered

Additional context

[runkecheng]

The init-sidecar container uses REPLICAS as an environment variable to create postStart script. After updating the number of cluster crd replicas, the spec.template of the statefulset changes.