[NBS] Checksum mismatch diagnostics improvements

[WilyTiger]

I want to distinguish between two situations:

1. Majority replicas with the same checksum:
   In this case, the scrubbing process has resolved the discrepancy through resync process. Also I want to add an option to enable a flag for copying data (this flag was introduced in [NBS] Add the ability to enable a buffer copying for write requests before sending them to replicas #2421). This situation may occur if a client does not wait for the write operation to finish and modifies the buffer.

2. Opposite situation:
   This requires closer investigation because scrubbing will not resolve such cases. Additionally, after [Blockstore] 2/3 reading for mirror-3 disk #2077, the client might not be able to read this block.