feat: Add non-streaming support to AI Assistant

[baumandm]

This PR allows toggling the streaming mode of the OpenAIAssistant. This is required when using a proxy that doesn't support streaming.

The default behavior remains streaming: true, but it can now be disabled via the model_args:

AI_ASSISTANT_PROVIDER: openai
AI_ASSISTANT_CONFIG:
    default:
        model_args:
            model_name: gpt-3.5-turbo
            temperature: 0
            streaming: false

I added the option into the config files to help document the feature, but it's completely optional and doesn't need to be specified at all.

When disabled, the LLM result is sent in single chunk when the response is complete.

[jczhong84]

@baumandm we're looking at the new version of Langchain which will have a much simpler way for the steaming, will try to cover the non streaming case when doing so