dataset issues

[mcfrank]

See: https://drive.google.com/drive/folders/1OZ1an_YzBEufWoPmMGvnkcNLBm0_lXyS?usp=sharing

README + CSV of issues. Also, code to flag issues + code to process videos, which is probably more important because I guess the data team will want to do that on future new videos too.

it would be nice to fix these issues upstream as I think it's probably important to provide users of the dataset with consistent video formats that are correctly rotated, and (if possible) allow them to pull a lower-resolution copy (e.g., 360x640) which is much faster to pull and faster for processing