Adding Pytorch/Tensorflow/Scikit-learn hyperparameter tuning problems

[teytaud]

We have real-world problems, based on MLDA or optical properties (ARCoating, Photonics).

We might want to include the hyperparameter tuning of Pytorch or Tensorflow or Scikit-learn models.

This can be done by reproducing the structure of e.g. ARCoating.

[teytaud]

This is done for Scikit-Learn (though it could be bigger).