Clustering of longitudinal interval-valued data via mixture distribution under covariance separability