Generalize composition diffusivity used for thermohaline calculations

[evbauer]

When the composition diffusivity enters into thermohaline calculations (e.g. when thermohaline_option = Brown_Garaud_Stellmach_13), currently we assume a H/He-dominated mixture. See https://github.com/MESAHub/mesa/blob/d215b38166687af0351c621f47ddf622415625d7/turb/private/thermohaline.f90#L128-L153

We already have all of the physics in the code to do the more general case elsewhere for element diffusion, so we should take advantage of that here.

[evbauer]

The current implementation is based on https://ui.adsabs.harvard.edu/abs/1993ASPC...40..246M/abstract

That's derived in the same general picture that we use for the full implementation of the Burgers equations for diffusion, but with some simplifying assumptions about composition and some fits used to estimate the Coulomb logarithm. So we should be able to translate the MESA implementation of https://ui.adsabs.harvard.edu/abs/2016PhRvE..93d3203S/abstract into an updated version of this without making any composition assumptions.