Add support for providing different base url for OpenAI llm and embedding model

[deshraj]

🚀 The feature

Requested by user San Nguyen on Discord: https://discord.com/channels/1121119078191480945/1125758905310519327/1193453420921765888

Motivation, pitch

It is possible that the embedding model might be running on different OpenAI account than that of LLM. Hence, we want to add support for configurable base url for both llm and embedding model.