`score_peak_file()` function in models.py should use `_group_preds_by_region()` to aggregate scores together, instead of using custom code