Add some forceeval and type double where possible

Thought it would maybe help with compilation

But not clear why this changes anything...

Fix missing inline

Added multipolar terms; doesn't compile on mac anymore

Appears to be working, additional testing required. Generalizing to handle G&V and G&T correlation integrals remains

No warnings from teqp code, all from externals

Very subtle problem with handling derivatives at zero density

Closes #38

Closes #40

Helpful for debugging

Tests should pass again

Still can't trace isotherms with mixtures yet

Ready for Python!

Using sigma in A seems to be necessary to avoid underflow in some numerical calculations, as is done in PC-SAFT

Still need the model and so on, but working

n[21] was -0.99218575e+1 instead of -0.99218575e-1