Determine a better approximation for mobility

[amjokisaari]

Currently, Hyrax approximates the atomic mobility of H in both phases as M = XD_alpha/RT. This works well in the alpha phase for up to X~0.05, but this is definitely far from the atomic mobility calculated as D_delta/dG_delta^2/DX^2. Mobility as calculated in Jokisaari et al, CALPHAD (2015) is not smooth and causes poor numerical convergence behavior. This formulation should be modified for a smooth mobility.