Next is the cross derivatives

Fix the parametric tracer; it was switching directions because the dot product check was not implemented properly.

TODO: add some tests to ensure this stays fixed, maybe P-R for CO2+ethane?

Added a mockup of how one would hardcode models into the shared library (a la CoolProp), dynamic loads just need to be replaced with hardcoded strings, probably encoded as binary

TODO: apply polishing in another variable if constant composition polishing fails

So fast...

It's slow. Quite slow.