Discriminative Speaker Representation via Contrastive Learning with Class-Aware Attention in Angular Space