Use of MeanOpacity in multigroup mode.

[RyanWollaeger]

Background

• Recently the MeanOpacity class in singularity-opac was extended to multigroup, where the opacity is discretized in frequency.
  • Monochromatic opacity can be treated in the new multigroup API through frequency-based accessors, which for a future MeanOpacity object could prescribe a sub-group frequency distribution to evaluate the opacity.
• It is worthwhile to have a true multigroup treatment for the multigroup mode in jaybenne, and make simplification where possible by removing the direct use of monochromatic singularity-opac variants.

Description of Changes

• Remove direct use of monochromatic singularity opacity objects.
• Employ MeanOpacity in frequency_type=multigroup.
• Add option use_opac_grps which currently resets the transport groups to the opacity groups, and also sets the scattering groups to the tabular absorption groups if absorption is pre-tabulated and scattering is analytic.

Note: scattering is currently always analytic; we will need an option similar to absorption to parse pre-tabulated opacity. In this scenario, if both absorption and scattering are tabular and use_opac_grps=true, the user would have to ensure the loaded tables have consistent group bounds, barring a remap step. If use_opac_grps=false and both opacities are tabular, the current changeset will use the particle frequencies to find a possibly different group through each opacity object (which should be okay).

TODO

• Use transport groups to set non-tabular opacity frequency array and number of groups.
• Reset to transport group structure to parsed tabular opacity group structure when using opacity tables (or not?)
• Use gmode in multigroup DDMC evaluations (see commented out gmode in struct).

Checklist

• New features are documented
• Tests added for bug fixes and new features
• (@lanl.gov employees) Update copyright on changed files

[RyanWollaeger]

This puts some jaybenne input accesses into the mcblock Initialization - thought this might be okay since mcblock as a package of course is only active when jaybenne is, and only uses input with jaybenne as the node label.

[RyanWollaeger]

Sort of interesting: switching from point evaluation to Rosseland group evaluation in the power-law scattering test seems to still pass (still match with the semi-analytic unit test at the tolerance), without any modification except for ensuring the MeanScattering gets physical temperature bounds.