jina-clip-v2: Multilingual Multimodal Embeddings for Text and Images