Fix LLM terms

[CharlesCNorton]

Updated all instances of mlm_client and mlm_model to llm_client and llm_model in the readme. The previous terms (mlm_client and mlm_model) are incorrect in the context of configuring Large Language Models (LLMs), as "MLM" typically refers to Masked Language Models, which is unrelated to the intended functionality. This change aligns the documentation with standard naming conventions for LLM configuration parameters and improves clarity for users integrating with LLMs like OpenAI's GPT models.

[afourney]

Indeed. This code dates back a while (to about the time of gpt-4-vision-preview), at which point it wasn't clear that to call the darned things, while making it clear the multimodality was necessary.

Anyhow: I agree with this, but changing this now will break things for people already experimenting with the library. Can we go in 2 steps. First let's enable the llm_ keywords as an alias to mlm_, and mark mlm_ as deprecated. Then we can remove mlm in a future release (e.g., 1 or two releases from now)