Need a mechanism to evict the AppliedManifestWorks from the managed cluster

[skeeey]

If the ManifestWork of one AppliedManifestWork is not found from the hub, the work agent will evict this AppliedManifestWork from the managed cluster after a grace period (the period has a default value, e.g. 10 mins, and it can be set by work agent argument)

After the AppliedManifestWork is evicted from the managed cluster, its owned resource will also be evicted from the managed cluster with kubernetes garbage collection. If one unfound ManifestWork is recreated after the grace period, the AppliedManifestWork of this ManifestWork will also be created again, this means, the resource for this ManifestWork may be deleted firstly, and then created again after a period.

The eviction will take place in the following scenarios:

1. The AppliedManifestWork hub hash is unchanged and its agent ID also is unchanged (e.g. the klusterlet is reconnected to the hub)
2. The AppliedManifestWork hub hash unchanged but its agent ID changed (e.g. the klusterlet is rejoined)
3. The AppliedManifestWork hub hash changed but its agent ID unchanged (e.g. the hub of one klusterlet is switched)

[skeeey]

/assign @skeeey

[skeeey]

code has merged, closing