Introduction to mathematical intuition behind ResNets.

[0xD4rky]

Hi Maintainers,

I have updated the ResNet documentation in Unit-2 (CNNs) and I have tried to include the following points in my documentation:

1. How "residual learning" allows networks to learn easier-to-optimize residual functions rather than directly mapping inputs to outputs.
2. How ResNets solve the degradation problem
3. How ResNets improve gradient flow by ensuring that gradients can propagate back through residual connections
4. How its possible to train "deeper neural networks" using ResNets.

Please have a look at my updated documentation and inform me if there is a need to better the documentation changes that I have done. Looking forward to it!