Refactoring of Lab AI for publishing

[HanaeRateau]

This PR prepares the lab to be published and then referenced in the optimization lab

TODO:

• (Hanae) Add the dataset explanation from lab optimization
• (Hanae) Explain the training
• (Manuela) Explain the evaluation with and without SOFA
• Start with scikit-learn to show what can be done with a simple MLP, then from scratch section, then pytorch
• Fix exercises numbering (start over in each section)
• (Hanae) Add introduction to PyTorch and Scikit-learn
• (Manuela) Overview should remind the pedagogical goals and maybe the requirements
• (Manuela) Lab Optimization: Add pre-requisites in overview with reference to Lab AI
• Make the user generate a dataset and use it in the lab Emio.lab_optimization_firstorder#5
• update buttons syntax
• Revamp the from scratch section. It needs way more explanations:
  • equations of the neurons gradients
  • give notations
  • Explain what we want to do exactly during the training (optimize the weights using the backpropagation)
  • Explain the backpropagation and chain rule from the loss function to the input layer
• Isolate the solutions