Function convert

Do the heavy lifting on the given subset (eg 'train')

(
    subset: str,
    clevr_box_path: Optional[Path],
    refclevr_path: Path,
    scene_path: Path,
    output_path: Path,
    no_caption: bool,
    medium: bool,
    templates,
    next_img_id: int = 0,
    next_id: int = 0,
)