Update models of FRB scattering and width distribution

Currently, this is old .Scattering is a log-normal, with a very poor fit to data. This should be updated, we have strong evidence for a better distribution. One could even calculate associated likelihoods.

We also have a much better intrinsic width distribution estimate. At the very least, we can re-estimate fits to data. There is some old code for doing this that I could ressurect.