Fixes for PR 2693

[paulromano]

@church89 I did a bit of refactoring on the batchwise_pr branch for openmc-dev#2693. There were three major issues that needed to be fixed:

1. The MPI handling in _update_vec was not correct; the vector that gets passed in is a rank-local list of atom densities, so it only needs to be updated on the current rank (not communicated across).
2. The restart behavior also appeared to be wrong. Since a keff search is performed before saving the state data, it means that when we restart we don't need to perform another search and instead can just restore the root by applying it.
3. The examples for material density updates didn't make sense to me. Because it was searching on a multiplicative scaling factor, I'm not sure what the behavior would be. The correct thing to do in my opinion is to do a criticality search on an absolute density, not a scaling factor.

I tried to make other improvements/simplifications. Let me know if you have any questions about the changes!

[church89]

@paulromano thank you for the review and fixes, looks good to me. I'll run some local tests and incorporate the changes into the main PR