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

Couple of inlines are needed to avoid duplication of functions between compilation units

Add tests for all binary pairs
Add conversion script for HFC mixture xi and zeta

Also allow paths to be optional

Previously you always got Lorentz-Berthelot no matter what

Not yet fixed for mixtures