What would you like to be added:
In the point InterpretReplica of ResourceInterpreter, enable interpreting customized scheduling result as well.
Or implement a new point to do the same thing in ResourceInterpreter.

Why is this needed:
Consider about the two scenario as following:

1. We have some clusters with the same distribution weight and approximate physical resources, and a lot of workload with replicas = 1 (or the number of replicas can't be evenly divided by weight). We hope to schedule these "remain replicas" into different clusters equally. But up to now, karmada scheduler will only scheduling them into the same cluster if the clusters has the same distribution weight.
2. When we retain some changes of replicas for workload in member cluster (for example, our HPA-controller works independently in member clusters), we may hope to sync the changes to controll plane, so that we will always have the consistent global perspective, for which it won't disturb our workload no matter when we stop retaining.

For scenario 1, we may adjust the distribution algorithm to solve the problem, but it is not a general solution, and it doesn't work for scenario 2. We think it will be better to do something in ResourceInterpreter.