Replaces sys.argv for argparse in hyperparameter_tuning.py, making the CLI more friendly. Adds number of trials in Bayesian optimization as a parameter

hyperparameter_tuning.py now builds a dictionary with the best hyperparameters instead of returning a partially trained model

Restructured models.py and hypermodels.py. Now all helper functions and custom layers are imported from model_utils.py

Restructured models.py and hypermodels.py. Now all helper functions and custom layers are imported from model_utils.py

Refactors KLD and MMD as independent Layers in experimental main.ipynb, thus avoiding global variable usage. Proposes a model including both KLD and MMD as regularizing terms in the loss function