This will allow parsers to export the duration of the current frame being output, if known, instead of using AVCodecContext.frame_size.