Make GNN explainable (e.g., with SubgraphX)

[janetzki]

Goal

Additional visualizations explain how the GNN works. Motivation: explain GNN

Tasks

• [x] #22
• [ ] Read this tutorial
• [ ] Read this tutorial

[janetzki]

Closed because we can explain the single-layer GNN with simple methods (e.g., by printing the neighbors of false positives that contribute to the positive prediction).