Add polarizability of Gubbins and co-workers

[ianhbell]

One key part is that the successive substitution is only needed in the Pade approximants (not in $U_p$, though I don't understand why) and that makes it much easier, and Gray et al. work out the partial derivatives in Eq. 2.11 and 2.12, so you can get the contributions directly.

Start at $\mu'=\mu$, calculate $E'$ from the gradient of the perturbation expansion $\hat\omega^0$, then iterate a few times to get $\mu'$. Actually doesn't appear to be as horrible as I had thought.