add functionality to perform hyperparameter optimization (HPO) and repeated seed runs during training/fine-tuning [PR#23](https://github.com/gridfm/gridfm-graphkit/pull/23)