Support embedding with "instructions" Again

Nobody has claimed this yet.

### Feature request

Hello,

First of all, thank you for developing infinity, an excellent package dedicated to inference for embedding models.

I am opening this issue to request support once again for the instructions feature discussed in issue #34.

From my understanding, to utilize the method added in [UKPLab/sentence-transformers#2439](https://github.com/UKPLab/sentence-transformers/issues/2439) within infinity, it seems necessary to pass either prompt or prompt_name at [this point](https://github.com/michaelfeil/infinity/blob/154160c8af464134103649ab1ce22406713eef5b/libs/infinity_emb/infinity_emb/infinity_server.py#L349).

Considering that the recently updated MMTEB has introduced an Instruction Retrieval evaluation category and that many models are being researched and developed as instruction-based embedding models, I believe integrating this feature into infinity would attract more users to the package.

Feel free to correct me if I’m mistaken.

### Motivation

https://github.com/michaelfeil/infinity/issues/34

### Your contribution

If you can provide a rough concept of how this feature should be integrated into infinity, I think I could submit a PR myself.

Start at infinity_emb/infinity_emb/infinity_server.py around line 349 and review issue #34 alongside sentence-transformers issue #2439. Trace how embedding requests reach the model and determine the integration points for prompt or prompt_name. Done means instruction-based embedding requests are supported in infinity and the behavior is covered by the relevant tests.

Written by the indexing model from the issue text.