DOC Write tutorial "How to add a datafit? How to add a penalty? in skglm" about skglm

[Badr-MOUFAD]

I agree with @PABannier on having a hands-on tutorial on adding new datafit.
We can get inspired by the approach we followed to add Poisson or Gamma datafits.

[mathurinm]

Go for it !

[mathurinm]

We have https://contrib.scikit-learn.org/skglm/add.html with a link in the header but if you did not find it, we may need to think about a better way to emphasize it. Feel free to update the content itself too, now that we have progressed on solvers and API

[PABannier]

I was thinking about a more hands-on tutorial, like adding the Poisson datafit. We start from the formulation of the datafit, and derive all the necessary quantities and questions one would need to ask itself to add the datafit (e.g. is my datafit Lipschitz). A mix between maths and code actually.

[PABannier]

Also, I find the raw_grad and raw_hessian names confusing. I think we should explain into more details what these two methods do in an example combining maths and code.