Replace finite differences with analytical derivatives in Brosh-Calphad EoS

[bobmyhill]

The Brosh-Calphad EoS currently uses finite difference to calculate dTheta/dP and the isothermal bulk modulus. For materials with a high bulk modulus, this is inaccurate (and there is no value of deltaP which would be suitable for the range of geologically interesting bulk moduli).

We should replace the finite difference schemes with analytical derivatives in the Brosh-Calphad EoS.

[bobmyhill]

Addressed by #645