Proper handling of write errors when a node contains multiple vdisk replicas

[ikopylov]

After the request #671 will be merged, we start supporting multiple vdisk replicas on the same node, but this support is limited in terms of error handling. If write to the node with multiple vdisk replicas ends with error, then only one copy in aliens will be created, so the number of copies can be less than specified quorum value. Also, if writing to one of the replicas on the node ends with an error, but the rest goes well, then there is no way to pass this information to the cluster level.

We should implement error handling properly with passing all required data to the cluster level (https://github.com/qoollo/bob/blob/master/bob/src/cluster/quorum.rs). This may require changes to the gRPC protocol, and may also require changes to the logic for placing of the aliens on disk