On documenting the methods of every distribution

[ricardoV94]

This issue of how/whether we should add extensive documentation to all the logp/ logcdf/ moment/ rng_fn functions of every distribution comes up now and then:, #5297, #4859, #5308 (comment), #6280

My take is that manually documenting of these methods has negative utility, since these docstrings are very verbose/repetitive and annoying to fix.

Also these methods are not user-facing, since it's not always intuitive how to transform the inputs into the form expected by them (e.g., call at.as_tensor_variable(inp) and transform them if needed). Most of our users are not expected to know how Aesara works, but without knowing that, trying to use/read/fix the implementations is not trivial. No reasonable amount of docstrings in the logp method is going to bridge that gap.

At most we could add logp(value: Variable, param: Variable) -> Variable if that seems helpful

For reference, scipy which is in a very similar position, does not seem too concerned with documenting their methods internally: https://github.com/scipy/scipy/blob/main/scipy/stats/_continuous_distns.py

https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.gamma.html