DOC L1-regularization parameter tutorial

[wassimmazouz]

This PR aims to add documentation about the mathematical context behind the choice of L1-regularization in skglm as it is one of the frequently asked questions.

Tasks :

• Created .rst tutorial and added its tag to the tutorials.rst file
• Introduced the problem and aim of the tutorial
• Finished the theoretical background
• Illustrated with an example
• Added references

[Badr-MOUFAD]

Thanks for the PR @wassimmazouz

I will follow along with you so ping me when you are done or need a hand

[mathurinm]

@wassimmazouz almost done, can you look into this rendering bug for the align environment that I was not able to fix ?

[wassimmazouz]

@mathurinm Is it a local issue? It renders correctly for me

[mathurinm]

We have an action that builds the doc on the PR and publishes it (build doc artifacts on circle CI), see the buggy rendering here too: https://output.circle-artifacts.com/output/job/c0047b51-706d-4b6c-ba96-4160db844411/artifacts/0/dev/tutorials/alpha_max.html#theoretical-background (from 1 commit ago)

[wassimmazouz]

@mathurinm it might have to do with the sphinx or mathjax versions. Some forums report similar errors in older versions of these packages. I tried to change the version and test locally but no progress for now.
I added a eqnarray to try if it works without changing anything else.

[Badr-MOUFAD]

The problem seems to appear only for multiline equations.
Fixed it by adding :nowrap: directive and writing plain tex.

Perhaps, sphinx/mathjax has troubles parsing the equation 🤷‍♂️

[mathurinm]

@wassimmazouz can you report this back to sphinx with a minimal example ?

[mathurinm]

Looking great, thanks @wassimmazouz @Badr-MOUFAD