Sorting configs and documentation

[sfluegel05]

Currently, we have a lot of configuration files (especially in the model folder) and a lot of different model implementations / configurations.

It would be nice to have an overview for future users what is implemented and how it should be used. I would suggest the following:

1. A "master example" giving the command and configuration files to train the best model architecture with the best dataset we have (probably GAT-v2 with augmented architecture and node properties + ChEBI50).
2. A short description of the parameters a user can sensibly change (i.e., swap between GAT / Resgated ConvNets / GATv2, number of layers, random node initialisation, ...)

Both should fit in the README.