Towards Multigroup API

[pdmullen]

This PR and PR summary text used generative AI.

Add photon multigroup absorption & scattering API

Adds a multigroup radiation API for photons by unifying multigroup support into the existing mean-opacity classes
(absorption and scattering) rather than adding new types. With ngroups=1 and bounds [0, ∞) these reduce exactly to
gray opacities; ngroups>1 enables multigroup transport.

• Group coefficients on (rho, T, group), built from a monochromatic model + group bounds or loaded from Spiner/SP5
  tables
• Half-open group bounds [nu_g, nu_{g+1}) with lower/upper tail groups (0. and IEEE +inf); frequency-entry APIs
  (GroupOfNu, ...FromNu); non-CGS wrappers preserved
• Shared helpers factored into mean_photon_utils.hpp; removed leftover ASCII loading

Full suite passes with tests + HDF5 on (686 assertions, 36 cases). Deferred to follow-ups: neutrino multigroup,
integrated multigroup emissivity.

[Yurlungur]

@pdmullen is this ready for review?

[RyanWollaeger]

Looks like test_mean_opacities.cpp got modified on main - it should be straightforward to resolve that branch conflict.

[RyanWollaeger]

Since the gray MeanOpacity class has been extended to multigroup here, does it make sense to support overloads of the opacity accessors that don't take a group index in the event that there is only one group (i.e. the tables are gray)? Or should that be a concern of the downstream codes?

[Yurlungur]

Some nitpicks but my comments are non-blocking.

[pdmullen]

Since the gray MeanOpacity class has been extended to multigroup here, does it make sense to support overloads of the opacity accessors that don't take a group index in the event that there is only one group (i.e. the tables are gray)? Or should that be a concern of the downstream codes?

These exist in this changeset already, I believe. However, I did have some concern that someone might accidentally use them with a multigroup opacity (ngroups > 1) which would result in the behavior of accessing the zeroth group, not a true frequency integrated opacity. Therefore, I added a PORTABLE_REQUIRE in 33103fb such that a user can only make these "gray" calls with a single group (i.e., gray) opacity.

[Yurlungur]

👍 merge when ready

[RyanWollaeger]

LGTM - one minor residual question, but feel free to ignore.