Integrating direct and indirect views for group recommendation: An inter- and intra-view contrastive learning method