Ensure that implicit complement cells appear last in DAGMC universes

[paulromano]

Description

The bug that this fixes is described in lots of detail at svalinn/DAGMC#934. It's debatable whether this is a DAGMC or OpenMC bug, but in either case, this PR will fix the behavior. Namely, for some DAGMC models, using MOAB 5.4+ currently results in particles incorrectly being found in the implicit complement volume when they shouldn't be. The change here ensures that all other volumes are checked first.

@pshriwise I don't have a test for this because I can't think of how one would do it. Let me know if you have any ideas.

Checklist

• I have performed a self-review of my own code
• I have run clang-format (version 15) on any C++ source files (if applicable)
• I have followed the style guidelines for Python source files (if applicable)
• I have made corresponding changes to the documentation (if applicable)
• I have added tests that prove my fix is effective or that my feature works (if applicable)

[pshriwise]

Thanks for tackling this @paulromano!! 💚

The shift in the EntityHandle sequence is really something we should test for in DAGMC, so I wouldn't worry about testing that.

Here in OpenMC, we could create a C++ unit test that loads one of the DAGMC files, creates a DAGUniverse object, and then verifies that the implicit complement index is at the back of the DAGUniverse::cells_ attribute.

[paulromano]

In addition to fixing the underprediction in k_eff that is discussed in svalinn/DAGMC#934, I'm happy to report that this also eliminates a lot of plotting artifacts that I (and others) have seen when using MOAB 5.4+. For example, here is a sample plot from the MSRE CAD model:

Before this PR

After this PR

[shimwell]

Confirming this makes nicer plots of the openmsr
Before

After

[paulromano]

@pshriwise I just updated one of our existing DAGMC tests to check for the implicit complement cell ordering.

[pshriwise]

Looks great! Thank you @paulromano.