can deal with heteroskedastic errors for deconvolver

[willwerscheid]

pass x/s to wrapper and then re-scale the output

[willwerscheid]

This is slightly annoying because we can't just re-scale the prior g. What to do? Ideally you'd want to be able to pass in x and s and initialize g at the solution and get the same result back, so you'd need to make it clear that g is a prior on the z-scores not (x, s). It might be easier just to tell the user to do their own scaling.