On the goodness-of-fit procedure for normality based on the empirical characteristic function for ranked set sampling data