Need of aiohttp based client support across different providers supported by litellm.

[rachitchauhan43]

In the context of recently uncovered performance issues with httpx, litellm team has released the fix to use aiohttp based client for OpenAI models. Great job on that

This same support, of using aiohttp based client, is needed across all providers that use httpx based client today in litellm so that all those provider integrations can be performant.

Open questions:

1. What is the plan to extend support to enable supporting aiohttp clients ?
2. Is there a plan to have some baseline perf test be a part of CI pipeline so that on every PR merge we know that performance is not compromised ?

[ishaan-jaff]

here is the plan @rachitchauhan43 #7544

[ishaan-jaff]

Is there a plan to have some baseline perf test be a part of CI pipeline so that on every PR merge we know that performance is not compromised ?

Yes - definitely, we already run load tests on every release. We will increase the expected RPS thresholds for each load test