support different LLMs for different agents

[zhangzaibin]

This PR is to fix issue #15

This PR includes the following modifications:

1. During the inference phase, the agent_id is passed to the inference_manager, where a distribution mechanism is designed to assign different ports to different agents, enabling hybrid LLMs simulation.
2. In the test folder, I have added testing functions and a test project for the OASIS inference system. Actual tests show that the parallelization design significantly reduces inference time.
3. Other minor fixes.

[yiyiyi0817]

@zhangzaibin Hello, thanks for the contribute and I am reviewing the PR. A preliminary advice is that could you please fix the format check first? Reference in https://github.com/camel-ai/camel/blob/master/CONTRIBUTING.md.