domain to rank map is wrong when domain ids aren't there

[cyrush]

this logic is wrong: https://github.com/LLNL/conduit/blob/5aef054fa682d7aa0ff307f7d034f16505f8991c/src/libs/blueprint/conduit_blueprint_mpi_mesh.cpp#L189

We need counting + offset b/c of multi domain.

[nselliott]

The current logic is that if it's multi-domain, there must be a state/domain_id, and that value overrides the par_rank value. If state/domain_id isn't there, it's assuming it's not multi-domain. This might be building in too many implicit assumptions, though.

[cyrush]

thanks @nselliott, got it -- at the moment we are hunting a mystery related to domain ids that are present, but I noticed this and I think we should upgrade the logic to be more robust.

[cyrush]

Also, to be clear -- the issue with domains ids is not a conduit problem, mystery lies elsewhere -- we are just using this method to help us as part of the debugging process.