Introduce deleteOption for GracefulEvictionTask

[Poor12]

What would you like to be added:
For application migration, we provide three ways, one is Never which means that Karmada will not delete the application in the failed clusters. But for the current GracefulEvictionTask, we cannot distinguish that GracefulEvictionTask should be kept or deleted when timeout is reached. In that case, we may need an flag to mark it.

Why is this needed:
Distinguish that GracefulEvictionTask should be kept or deleted when timeout is reached.

[zishen]

Before answering and looking into this case, I might need to ask some questions to understand this.

1.How long does this failed cluster last?
1.Can the value of unschedulable-threshold be set higher, until the failed cluster recovers?

[Poor12]

Another way is use Producer to judge it. For example, if a GracefulEvictionTask's producer is from taintManager, we may judge it from Failover.Cluster.PurgeMode and if a GracefulEvictionTask's producer is from applicationFailoverController, we may judge it from Failover.Application.PurgeMode.
I prefer to the former one.
/cc @RainbowMango for opinions.

[Poor12]

Before answering and looking into this case, I might need to ask some questions to understand this.

1.How long does this failed cluster last? 1.Can the value of unschedulable-threshold be set higher, until the failed cluster recovers?

Before answering the question, I'd like to recommend you some documents about multi-cluster failover. You can get some inspiration. :)

[zishen]

Before answering and looking into this case, I might need to ask some questions to understand this.
1.How long does this failed cluster last? 1.Can the value of unschedulable-threshold be set higher, until the failed cluster recovers?

Before answering the question, I'd like to recommend you some documents about multi-cluster failover. You can get some inspiration. :)

sorry, my description may not be accurate. But I did know the failover.
I had got the k8s will delete the fault pod when node failed in 1.26. You can see here Kubernetes 1.26: Non-Graceful Node Shutdown Moves to Beta
So, I want to know why need keep the application in the failed clusters, and how to do it.

[Poor12]

For multi-cluster failover, some users may hope that application can remain on the failed cluster, and after the failure is restored, it is up to the user to decide how to deal with the redundant copy. In this case, we may want to keep the GracefulEvictionTask.

[zishen]

For multi-cluster failover, some users may hope that application can remain on the failed cluster, and after the failure is restored, it is up to the user to decide how to deal with the redundant copy. In this case, we may want to keep the GracefulEvictionTask.

ok，I got it. It looks good.
Thanks for your patience.