[Feature Request] Implement graph normalization

## 🚀 Feature
Implement graph normalization function based on papers:

1) [GraphNorm: A Principled Approach to Accelerating Graph Neural Network Training](https://arxiv.org/pdf/2009.03294.pdf)
Already uses DGL: https://github.com/lsj2408/GraphNorm/blob/master/GraphNorm_ws/gnn_ws/gnn_example/model/Norm/norm.py
https://github.com/microsoft/ptgnn/blob/master/ptgnn/neuralmodels/gnn/messagepassing/graphnorm.py

2) [LEARNING GRAPH NORMALIZATION FOR GRAPH NEURAL NETWORKS](https://arxiv.org/pdf/2009.11746v1.pdf)
Already uses DGL: https://github.com/cyh1112/GraphNormalization/blob/master/norm/graph_norm.py.

They are regular normalization: x - mean / std, but applied to the whole graph. The main difference in that (1) introduces alpha:

## Motivation

Results from papers:
1)

2)

## Additional context

I am sorry if this is already implemented. Couldn't find it mentioned in the website, or github or in the code of GIN.