Fix/test virial coefficients for all models

[ianhbell]

Due to a problem with autodiff (autodiff/autodiff#351) some virial coefficient calculations are incorrect when you have terms like pow(rho, int) and rho is evaluated at zero. Thus all models should have their virial coefficients compared with a calculation at a very low, but nonzero, density

[ianhbell]

Actually a lot of models are hit by this bug. Many can be worked around with the use of powi instead of pow