There are some problems to solve still:

1. I had to name new object nc_galaxy_wl_likelihood in order to not mess with previous naming scheme. I don't think that's the best name.
2. nc_galaxy_wl_likelihood currently has 2 methods eval_m2lnP and kde_eval_m2lnP for each of the methods. I don't know if that's the best way to present these methods. We can use a flag maybe?
3. I still have to implement the likelihood algorithms, both for KDE and integration (we should discuss the integration algorithm you mentioned @vitenti).
4. If we wish to use different source positions priors and descriptions (i.e. using coordinates, $r$ or $\theta$) it might be best to make nc_galaxy_sd_position contain other two objects describing image position and redshift separately. This may not be necessary if we're not dealing with miscentering.
5. I think it's worth considering adding a way to deal with shape measurement error. Maybe by creating an optional object nc_galaxy_sd_shape_m_err witch could be contained by nc_galaxy_sd_shape.

Also, I'm already working on unit testing. That should be up today still.

coverage: 68.063% (+0.7%) from 67.332%
when pulling 073909e on wl_rebuild
into 192c5e5 on master.

I can't create unit tests yet since I haven't implemented nc_galaxy_sd_*_new () and thus can't create neither nc_galaxy_sd_* objects nor nc_galaxy_wl_likelihood objects by extension. Maybe we merge the PR without unit testing while I work on the implementation of basic nc_galaxy_sd_* objects?

Also, I need some directions on how to write these new objects on numcosmo-docs.sgml.