Pyxtal gnn proxy

[carriepl]

Left to do :

• Implement surface cutting (will generate multiple possible slabs so we might need to define a hyperparameter saying how many of those we will randomly sample and take the min of)
• Select an adsorbate and add it to the graph
• Ensure that the GNN library xists and can be loaded
• Load a GNN checkpoint
• Run the model on the generated graphs

[carriepl]

This PR is closed here and re-opened on the private repo.