Fix multi view model prediction

[Dominic-DallOsto]

Fixes #327

I changed the find_videos_files_for_views function to return a list of lists of videos like [[session0_view0.mp4, session0_view1.mp4, ...], [session1_view0.mp4, session1_view1.mp4, ...], ...].

This then lets us loop over all sessions in the training videos directory and feed the videos from all the different views to the model.

Additionally, I modified litpose predict for multi view models to accept a list of videos from multiple views, or a list of directories that each contain videos from multiple views. If there's a different way of handling this that you prefer that's no problem!

[Dominic-DallOsto]

Ahh, the test for find_video_files_for_views is failing because I changed its behaviour. Happy to update it if you think the new behaviour makes sense.

[themattinthehatt]

Hi @Dominic-DallOsto thanks for the bug find/fix! I just returned from traveling so I'll take a closer look at this later this afternoon
@LennyAharon

[ksikka]

Thanks for the feature! I liked the proposal and just made a few minor tweaks as I reviewed.
Dominic-DallOsto#1

(My first time reviewing a PR by following up with a PR - I thought this might be the easiest way to minimize the work for you to accept the changes)

Do you mind adding test cases for the new split function?

[Dominic-DallOsto]

Perfect, thanks! I added some tests. Let mw know what you think

[Dominic-DallOsto]

I also tested this with my multi view model and it works with the nice pretty printing now