Make Ansari LLM Generate a Non-Streamed Response

[OdyAsh]

Currently, the ansari.py is coded in a way that always anticipates litellm to return a streamed response (e.g., many "yield" statements, etc.).

We want to refactor and update this code so that it incorporates non-streamed generation as well (which can be used with certain endpoints like WhatsApp, where a message has to be sent as a whole and can't be streamed, etc.).

Additionally, this seems to be affecting the final response time to WhatsApp users. See this issue for more info.