Add supports for weighted edge sampling

[vincentX3]

🚀 The feature, motivation and pitch

In large graphs with more than ten million edges, the need for edge sampling is inevitable. Hopefully, GLT can support sampling based on edge weights, so as to better utilize edge features.

Related Content:
PyG

• Implement weighted random walks rusty1s/pytorch_cluster#140
  DGL：node2vec random walk and top-k sampling
• https://docs.dgl.ai/en/0.8.x/api/python/dgl.sampling.html#neighbor-sampling

Alternatives

No response

Additional context

No response

[husimplicity]

CPU version already merged #72. You're welcome to try CPU sampling and GPU training :) @vincentX3