Projector can miss few partitions of an index during file transfer rebalance

It has been observed in a customer setups that during rebalance, after addInstances request, projector will miss few partitions. Any mutations on those instances will reach indexer as UpsertDel() with empty keys and this can lead to corrupt indexes.

The issue is seen only for partitioned indexes. For non-partitioned indexes, the issue does not exist. Also, the issue is seen only with file based rebalance. DCP rebalance does not encounter this issue