Allow chi_effective as a fit parameter

We should support chi effective and possibly also chip in the fit parameters. This will require defining a transformation for the injections and a jacobian for the initial prior.

@jacob.golomb can you take this on? I'll need to add you as a project member to assign issues to you.