Add more examples:

[DerThorsten]

We should add more examples for the following things:

• infernos transformation pipeline, show / highlight the difference between torchvision and infernos transformations
• usage of trainer with non-trivial dataset (smth where num_input>1 and num_output > 1)
• load a model which was trained/saved via inferno and use this trained model to predict
• ...

[bstriner]

@DerThorsten on point 2, I think num_output > 1 is buggy and needs some tests added.