[example] multi gpus training for mag240m

[maqy1995]

Description

the code is similar to https://github.com/dmlc/dgl/blob/master/examples/pytorch/graphsage/train_sampling_multi_gpu.py

related discussion:https://discuss.dgl.ai/t/will-mag240ms-baseline-code-use-multi-gpus-training-bring-efficiency-improvements/1878

Points that may need attention:

1. It may be necessary to perform performance testing again on the DGL server. The first Epoch may be slower because the feature needs to be cached.
2. I did not adjust the parameters (such as learning rate), which may cause the final accuracy rate to change.
3. Maybe the README should be updated.

Checklist

Please feel free to remove inapplicable items for your PR.

• The PR title starts with [$CATEGORY] (such as [NN], [Model], [Doc], [Feature]])
• Changes are complete (i.e. I finished coding on this PR)
• All changes have test coverage
• Code is well-documented
• To the my best knowledge, examples are either not affected by this change,
  or have been fixed to be compatible with this change
• Related issue is referred in this PR
• If the PR is for a new model/paper, I've updated the example index here.

Changes

[BarclayII]

Thanks a lot!