Video Manipulations Beyond Faces: A Dataset with Human-Machine Analysis