Add support for embedding model prompt prefixes

Some embedding models like
https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1
and
https://huggingface.co/nomic-ai/nomic-embed-text-v1.5

Require special prompt prefixes for use cases like RAG to work better, so we need two more fields in the model configuration for injecting these prefixes: one for index and one for query.