[quiver-team] Add Quiver example for multi-gpu training using GraphSage

[eedalong]

[quiver-team] Add Quiver example for multi-gpu training using GraphSage.

[luomai]

Hi PyG team,

This is the PR to contribute PyG examples with GPU-based sampling + feature aggregation on a GPU and multiple GPUs through Quiver. We observe substantial speed up in the PyG examples with GraphSage.

We will create another PR for distributed sampling + feature aggregation in 2 - 3 weeks. We have been testing the performance of distributed sampling in the last few weeks, and the code will be ready soon.

Quiver team.

[rusty1s]

I think this looks great. Thanks for the efforts. It might be also a good idea to link to this example from the README.md. WDYT?

[luomai]

I think this looks great. Thanks for the efforts. It might be also a good idea to link to this example from the README.md. WDYT?

@rusty1s Yes, this a good idea! Where shall we put this link? Maybe under the "Implemented GNN models"?

[eedalong]

I think this pr is ready to merge, what you guys think? @luomai @rusty1s

[rusty1s]

Thank you very much! This is great.