Flexible dataloader for evalutation scenarios

[tdrose]

For the evaluation of the project, we discussed a few scenarios from simple to more complex. Dataloaders (training, validation, and testing) for each scenario should be implemented.

• [x] One cohort, randomly leave out molecules
• [x] One cohort, leave out molecules, such that they never occur in one dataset
• [x] One cohort, leave out complete datasets
• [x] Diverse datasets, leave out random molecules
• [x] Diverse datasets, leave out molecules, such that they never occur in one dataset
• [x] Diverse datasets, leave out complete datasets

[tdrose]

Just one scenario missing, otherwise the new data(loader) framework is working.

[tdrose]

Dataloader for small image patches required. This can be done e.g. with RandomCrop. But It requires a different DL framework, since with the current symmetric images, it will result in many empty patches because they are off sample. So I need to load and crop the images independently.