Refactor for labels & image outlier

[matanper]

Before implementing label & image outliers checks, I needed a few changes:

1. easy way to get for the samples I run on the real dataset index. Will now be able to do in the update:

properties_with_id = []
for index, image in enumarte(batch.images):
  property = func(image)
  dataset_index = batch.get_index_in_dataset(index)
  properties_with_id += [(dataset_index, property)]

2. Made label & prediction properties to either return flat list or list of lists
3. Made it possible to run checks on VisionData classes that don't have labels. Until now it would have raised exception when trying to update the VisionData cache, so added try/except

[JKL98ISR]

also other functions missing type hints i.e get_index_in_dataset