Previously we would scale the bitmap size/position to a proportion
of the original video frame, then scale it back up again to the DCP
container. This didn't take into account some cropped cases where
the picture would end up the same shape but the subtitles would be
stretched.