[keep-balance] replication level distribution should account for volumes' built-in replication
In olden days, the replication level distribution (logged by keep-balance after each run) reflected the underlying storage device's replication.
This was changed in #15305, perhaps inadvertently, while fixing the byte counters.
We should restore the old behavior, so if 2 blocks are stored on 1 volume that has underlying replication 3, and 2 other blocks are stored on 1 volume that has underlying replication 1, the logged distribution looks like this:
0: 0 1: 2 2: 0 3: 2