feature: initial framework for training and evaluating models

[saraheisenach]

This PR adds the initial framework for training and evaluating ML models. The changes in this PR include:

• Framework to run model training from the command line using scripts/main.py and configs/config.yml
• Registry to register and load class names for trainers, models, and tasks when running from config file
• Data preprocessing to create data.pt file
• Model, Trainer, Evaluator, Scheduler, and Task classes
• CGCNN model class
• Logging per epoch for training run

[shuyijia]

Tested dev branch on PACE; was able to successfully generate data.pt file and run training on CGCNN for 10 epochs.

[shuyijia]

Tested dev branch on PACE; was able to successfully generate data.pt file and run training on CGCNN for 10 epochs.