Can you please make an end-to-end example of a heterogeneous link prediction task using most of high level dgl features?

## 🚀 Feature

I wish you can make an example of a link prediction task using most of high level dgl features including EdgeDataLoader, heterogeneous data, dgl built in convlayers (not novel convs)?
Altogether, ```end-to-end heterogeneous link prediction task using EdgeDataLoader and DGL convs```
## Motivation
Now there are link prediction examples in example folder, but they either dont use EdgeDataLoader (because maybe dataset is small enough), or make their own MGFs. With user guide, it is not an end-to-end but snippets of relevant code.
And heterogeneous graph link prediction must have a few lines of code to adjust, but mostly in example of DGL, it is only for hetero node classification. There is no single end to end hetero link pred example using ```EdgeDataLoader``` and ```extensive only DGL builtin simple model``` to be intuitive for all.

So I suggest to make one as above, ```end-to-end heterogeneous link prediction task example using EdgeDataLoader and DGL convs```

## Alternatives

## Pitch

## Additional context