add jacobians as transformations and implement ability to change variables when computing PDFs

See LIGO-T2100136 for many useful formula.

We may want to implement a class that eats "standard" distributions and converts them into distributions over different variates by applying the appropriate Jacobians. This should be extremely flexible, but could be complicated as we need to be able to perform the inverse mappings from "new" variates to "old" variates in order to call the "eaten" class's prob, logprob methods. Sampling should be a breeze, though.