Improve handling of "Analytics collections in different partitions have different DCP states"

Currently we fail any rebalance once we encounter a dataset whose partitions do not have identical state. This should not be necessary:

1. On a rebalance-in, we could simply skip any dataset(s) that we encounter this state. Once the states are in sync, the subsequent rebalance will rebalance the problematic dataset(s). This can be accomplished very quickly (unfortunately a bit late for 7.2.0).
2. I think we could allow the mismatched datasets to be rebalanced, reverting all DCP state to the earliest copy. We can then catch up any out-of-sync dataset(s) post-rebalance.
    
   EDIT: will open another issue to consider #2 and other improvements- tracking #1 with this issue.