After hpa controller determines the replica nums needed per cluster, it also
controls the distribution of target objs (rs or deployment) into the correct
clusters by telling the corresponding controllers, which clusters they should
put the objects into (passed as list of selected clusters in annotations).
