Cross-Modal Contrastive Learning for Domain Adaptation in 3D Semantic Segmentation