feat: Add INSTRUCTOR Embedder (v2)

[awinml]

Code Migrated from deepset-ai/haystack#5836.

Related Issues

Fixes #5242

Proposed Changes:

Add support for the INSTRUCTOR family of Embedding Models. They tailor the embeddings for different tasks and domains using a prompt (instruction) for each embedding.

• The implementation is based on the Embedders Proposal and Embedder Design (#5312).

• Adds InstructorEmbeddingBackend, responsible for performing the actual embedding computation, implemented as a singleton class in order to reuse instances.

• Adds InstructorTextEmbedder, a component that embeds a list of strings into a list of vectors.

• Adds InstructorDocumentEmbedder, a component that embeds a list of Documents. The embedding of each Document is stored in the embedding field of the Document.

Since the INSTRUCTOR Embedding Models are a modification of the Sentence Transformers models, the implementation is similar to the Sentence Transformers implementation (#5567).

This code was written collaboratively with @vrunm.

How did you test it?

Unit Tests have been added for each component.

[ZanSara]

@anakin87 note that CI is not running, so let's run the tests locally at least before approving. We'll take care of the CI separately.

---

Maybe that's because this file is missing the .yaml extension 👀 https://github.com/deepset-ai/haystack-extras/blob/main/.github/workflows/components_instructor_embedders

[anakin87]

(@dfokina can you have a quick look at the docstrings?)

I felt free to make some minor changes, but
kudos again for a great job --> @awinml and @vrunm

It would also be nice if you open a PR in https://github.com/deepset-ai/haystack-integrations,
to showcase these new INSTRUCTOR Embedders, so that they can also appear on this web page.
Feel free to ask for help (if needed). 💙

[awinml]

@anakin87 Thanks!

The changes look good. I will open a PR in https://github.com/deepset-ai/haystack-integrations with examples to showcase these new INSTRUCTOR Embedders.