Description of problem:
CpuOverCommitDuration not working correct in cluster policies, also for Power_Saving and also for Even_Distribution policies, it's mean that if I have host with vm, that CPU of host out of define thresholds and I also have host that CPU of host in given bounds, vm start migration without waiting given in CpuOverCommitDuration field time.

Version-Release number of selected component (if applicable):
is23

How reproducible:
Always

Steps to Reproduce:
1. Add two hosts to engine, add vm and run on one of host, change cluster policy
or to Power saving or to Even Distribution, I will describe reproduce process for Even Distribution, by default CpuOverCommitDuration is 2 min and HighUtilization on 80
2. Load host CPU with running vm to 85%(you also can try some script, to run infinitive loop) 

Actual results:
Engine not waiting two minutes, but begin migration to host without loading CPU much earlier.

Expected results:
engine waiting given CpuOverCommitDuration time and just after this start migration if it still actual.

Additional info:
whole process loading CPU and waiting until vm will migrate on other host, takes less than minute