Could also be helped by perfect branch prediction

The use of void* in the interface is not my favorite solution...

See https://github.com/usnistgov/teqp/issues/9

More testing is still needed to identify other issues

Bump version

Stopped iteration if dx values are all < 0

And to python interface

* Add ability to access a single departure term in a multi-fluid model

* Bump version to dev

Problem: compile-time time arguments remapped to runtime arguments, in future the runtime arguments should come after the model

* Add methods for calculation of pure fluid critical points

* Pass the template arguments along

* Expose the values and Jacobian

Closes #8

* Add partial molar volume as a possible output

* And expose to python