Computer Vision-Driven Movement Annotations to Advance fNIRS Pre-Processing Algorithms