DINOde: Continuous Vision-Text Alignment for Open-Vocabulary Semantic Segmentation