update replicas will cause all pods been rescheduled

[freeznet]

when updating replicas for a sink, all pods will be rescheduled, which is not expected behaviour. Should fix this BUG and add related tests.

[freeznet]

This is because the FunctionDetails have been passing as STS container template command and the FunctionDetails is changing when replicas changed.

[jiangpengcheng]

seems the parallelism is getting from the spec.replicas, since we are using k8s for auto-scaling, how about setting the parallelism to hardcoded 1?

[tpiperatgod]

I have a proposal to use a proxy service to provide the ability for pulsar functions managed by Function Mesh to interact with the cluster.

In a preliminary idea, this service should provide the following interfaces.

• Get the replicas number of the function
• Get the status of the function