Context model option doesn't show

[ksikka]

is_context_dataset logic makes the assumption that frame filenames DO NOT have any separators like - or _.

Given a like frame_303.png it will look for context frames like frame0303.png, not find them, and assume the dataset is not a context dataset.

Workaround is to name files like frame303.png instead.

[ksikka]

Also filenames must use the img prefix instead of frame, until the lightning pose version is upgraded to pick up paninski-lab/lightning-pose#216. Otherwise it will fail during model training.