Refactor the descheduler code base to allow a quick implementation of the proposal in https://github.com/kubernetes-sigs/descheduler/pull/1354.

DoD:

• extend the PodEvictor with a ResetCounters so it can be initialized only once outside of the descheduling cycle: https://github.com/kubernetes-sigs/descheduler/pull/1447
• generate UID for test pods: https://github.com/kubernetes-sigs/descheduler/pull/1454
• PodEvictor: turn an exceeded limit into an error: https://github.com/kubernetes-sigs/descheduler/pull/1456
• Thread-safe pod evictor: https://github.com/kubernetes-sigs/descheduler/pull/1458
• Refactor descheduler_test.go: https://github.com/kubernetes-sigs/descheduler/pull/1459
• e2e: TestRemoveDuplicates: limit the tested namespace: https://github.com/kubernetes-sigs/descheduler/pull/1471
• e2e: TestTooManyRestarts: run descheduler as a whole instead of a single plugin: https://github.com/kubernetes-sigs/descheduler/pull/1472