LLM as text encoder implementation from LLM2CLIP

[gabrieletijunaityte]

What does this PR do?

• Normalisation check moved to init
• Cosine similarity fixed not to normalise twice.
• The implementation of LLM as a text encoder from LLM2CLIP.

I had to do a lot of work arounds to make it work due to dependency issues and unclear environment recommendations from LLM2LCIP. Right now, it uses my forked LLM2vec repo and LLAMA customised version to get rid of flash-atto and spda.
What remains is to unify/decide on model dtypes and potentially adapt LLM2CLIP vision branch as eo encoder.

Before submitting

• Did you make sure title is self-explanatory and the description concisely explains the PR?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you list all the breaking changes introduced by this pull request?
• Did you test your PR locally with pytest command?

[vdplasthijs]

Great work, curious to see the results! I'll merge this, and we can change the normalisation in a new PR? Thanks!