Feature/Data Manager

• Creating global DataManager class that loads data manager specific to training plan types
• Creating DataManager for Torch and Sklearn based training plans
• Test/train split features for data manager
• Creating TorchTabularDataset class to auto-load datasets that are provided as pd.DataFrame or np.Array.
• Update notebooks to use Fed-BioMed DataManager instead of torch DataLoader
• Add preprocess handler that call proprocess functions just before training loop
• Unit tests for new DataManager, TorchDataManager, SkLearnDataManager and TorchTabularDataset
• Update default models for model validation