adding functionality to provide explicit validation dataset

[llevitis]

I'm adding functionality to provide a validation dataset directly instead of splitting training data in scenarios where data leakage needs to be avoided between training and validation. I've made the following changes:

• The fit function in base_model.py has been updated to accept an optional split_labels argument that is a list comprised of train and val entries to be used for splitting the input *data.
• Functionality has been added to dataloaders.py to split the input data either using the split_labels or by splitting data directly into 90% for training and 10% for validation