Models in "5.2 Edge Classification/Regression" do not use edge features for edge classification tasks, or the tutorials is unclear about it

## 📚 Documentation

Hi all, I have a question about the "5.2 Edge Classification/Regression" tutorial.

When creating a test graph, there is this line:
```python
edge_pred_graph.edata['feature'] = torch.randn(1000, 10)
```
Which *supposes* that edge features of the graph will be used for creating the model.

But examining the tutorial further, it seems as if the network does not actually use these edge features at all, but rather node features only? I'm pretty new to DGL, so I was wondering if they are still somehow used by graph networks under the hood. If not, this tutorial seems a bit misleading to me, even though it isn't that difficult to modify the example network to actually use edge features for prediction.