Verify whether comparison of errors for creating replication jobs works as expected

When testing whether we need to create replication jobs we compare errors returned by the Gitaly nodes in Praefect's coordinator: if the errors are the same, then we assume that all Gitalies did the same thing and don't schedule replication jobs. We see that we hit this error condition from time to time in production.

gRPC errors cannot be compared with each other directly, but that's exactly what we do in the coordinator. We should see whether this works as expected or whether this causes us to schedule excessive replication jobs which could be avoided.