fix: fix the race that we stop selecting a cluster as soon as its spec changes

[ryanzhang-oss]

Description of your changes

1. Fix the race that we stop selecting a cluster as soon as its spec changes since we compare the observed generation of the condition.
2. We also want to relax the eligibility of the cluster to not remove work from a temporality not healthy/available cluster. We can add more filters later when we introduce more complicated scheduling mechanisms.
3. Fix the bug that workload placement logic looks at the overall condition of the member cluster. It should only care about the work agent status
4. Fix the related bug on logics to determine if one should queue a member cluster to trigger CPR reconcile

Fixes #182

I have:

• Run make reviewable to ensure this PR is ready for review.

How has this code been tested

1. existing e2e was flaky, verified that it passes consistently
2. add UT

Special notes for your reviewer

We will eventually adopt the chain of filters in cluster selection just like in K8s.