Integrate `encode_query`/`encode_document` from SentenceTransformers `v5`

[Samoed]

Description of the feature

For now we're using same prompts as SentneceTransformers, but in future there can be different models that would use Router. I think we can simply pass task type based on PromptType to encode without any other changes

https://github.com/UKPLab/sentence-transformers/blob/cfb492fce85e142bca6a2cdeca6492d5548fe1c5/sentence_transformers/SentenceTransformer.py#L416

Doc ref https://www.sbert.net/docs/package_reference/sentence_transformer/SentenceTransformer.html#sentence_transformers.SentenceTransformer.encode_query

[tomaarsen]

Hello!

Yes, the encode_query and encode_document have been requested as shorthands that differ in exactly the two ways described in the documentation. You can still provide your own prompt_name or prompt to encode_query or encode_document.

The goal here is to provide a bit more unification in how models that differ across queries and documents are used, especially for third parties like MTEB that might benefit from that.

• Tom Aarsen