fix: preprocessing

[senpai-a]

1. Fixed a key error when reading the status of pandas.read_csv.
2. The landmarks path does not correspond to the output dir in the
   landmark extraction step in README. It's confusing to use beacuse this
   landmark path is actually a directory containing subfolders for landmarks
   of every video. In README its never mentioned the landmarks for each
   video should be in a folder whose name is the video name. Therefore, I
   changed the rule to find landmarks in extract_facial_regions.py. The
   user use a single dir to store all videos' landmarks and use this dir in
   cropping and alignment step.