Rework of PKs to split out parseParameterList, make SetEvaluator() call EnsureEvaluators()

[ecoon]

This PR is joint with amanzi/ats#270 and amanzi/amanzi#860.

Those PRs:

1. make it so that we don't need "source term is differentiable", as the PK can just ask the evaluator. (see should the sub-dag be valid as soon as an evaluator is required? ats#167)
2. This also potentially changes the order of which evaluators are created; as such some evaluators which create multiple variables (e.g. molar_density_liquid and mass_density_liquid), now the other variable is required first. There is currently no good way of dealing with this in Amanzi-ATS, so we're stuck changing the name in the input file. This is going to annoy people, but maybe we can just script the duplication of the list under both names in the input converter?