fix: fallback for calculation embedding dimensions if not available

[undo76]

Related to #43, where the number of embedding dimensions is missing, I am adding a fallback mechanism that embeds a dummy sentence in order to infer the number of dimensions if they are not available in the configuration.

[lsorber]

Thanks for the PR @undo76! I added a commit that moves the get_embedding_dim function to the _litellm.py module because I think it's a useful function that other modules may also want to use. I also took the opportunity to add a second model info function to grab the LLM's context size as well. Unless you have any comments, I'll merge when the pipeline succeeds.

[undo76]

Both changes look like an improvement, indeed. Thanks for merging them.