Add streaming support to text converse endpoint

[Ashesh3]

Frontend change: ohcnetwork/ayushma_fe#7

This pull request introduces crucial updates to the backend that enable real-time streaming support for Ayushma (AI) responses. The changes ensure that users can interact with the AI in a more engaging and dynamic manner, as tokens are displayed as soon as they are received from the AI.

• ChatOpenAI is initialized with streaming=True and with a callback_manager which receives each new token as it is received from OpenAI.
• Django's StreamingHttpResponse is initialized with content_type="text/event-stream" and the LLM's output is streamed to the client in the below format.
• Employed start_blocking_portal to initiate the lang_chain_helper.get_response coroutine, which retrieves the AI responses in real time and populates the token_queue accordingly.
• Each token is added to a queue in the StreamingQueueCallbackHandler which is then processed to create a streaming json output.
• Implemented a loop that continuously reads tokens from the token_queue, appending them to the chat_response string and yielding an updated JSON response for each token. The loop terminates when the RESPONSE_END token is encountered.
• Format of message: {"chat": chat_id, "delta": delta, "message": message, "stop": stop} where delta is the new token received and message is the entire message streamed so far.
• Minor fixes such as fixing a logical issue with split_text and refactoring the code to use ChatMessageType instead of hardcoded numbers are also made.

These backend enhancements work in tandem with the frontend updates to provide a seamless, real-time streaming experience for Ayushma (AI) responses, significantly improving user engagement and interactivity.