Add the Cross Entropy Loss

Descritpion

The Cross Entropy Loss is a classification loss that relies both on the Softmax function and the Cross-Entropy formula. It is a widely used loss, that is for example, choosed for the PyTorch reference benchmarks. The combination of the two formulas leads to great math simplifications, so CELoss is computationaly efficient.

More details can be found on the PyTorch page : https://pytorch.org/docs/stable/generated/torch.nn.CrossEntropyLoss.html