This requires separation of enthalpies into FluidEnthalpyData
and SolidEnthalpyData, hence the many changes.

It seems to avoid some issues after Eigen update with the SparseLU
solver crashing.

Now this test checks all molar fluxes, too.

From both phase transition models.

Reducing storage a little more.

The xnWG is not used in assembly and is required only
internally (and in the unit test) but is easily substituted
with 1-xnCG.