[proposal] The Coscheduling PreFilter method should not return Error if failed

[eahydra]

What is your proposal:

If some Pods with the same PodGroup CRD object are created and scheduling fails, the coscheduling PreFilter returns framework.Error to stop preemption, but this is not the best option. It should return framework.UnschedulableAndUnresolvable.

Why is this needed:

framework.Error is special in the scheduler. It represents an internal service exception. We can determine whether the scheduler is healthy by identifying whether there are Errors, but a plugin that frequently returns Errors will prevent us from identifying exceptions.

Is there a suggested solution, if so, please add it: