Add capabilities for decoding/decompression of images/videos

[KonstantinWilleke]

For datasets in original resolution, our current approach doesn't scale. We'll add encoded/compressed versions of images/video at export level and then add the functionality to read these files in the __getitem__

[pollytur]

for the sake of organisation - WIP by @tomschewski in https://github.com/tomschewski/experanto/tree/foundation_torchcodec

[schewskone]

Adressed in @ #81. Read PR comment for more context.