Examples/Tutorials for training TF GNN models using GPU based on sampling with graphscope.learning

**Is your feature request related to a problem? Please describe.**

Obviously, it is supported, but it would be nice if we could have some tutorials/notebooks to show the capability.

It doesn't require much engineering effort but does need some crafting and testing work on existing GNN tutorials.

**Additional context**
Add any other context or screenshots about the feature request here.