An agent implementation of MathChat

[yiranwu0]

Why are these changes needed?

Implement MathChat using the agent framework.

Related issue number

See tasks in #1066. The blog post about MathChat will be written in a separate PR.

Checks

• I've included any doc changes needed for https://microsoft.github.io/FLAML/. See https://microsoft.github.io/FLAML/docs/Contribute#documentation to build and test documentation locally.
• I've added tests (if relevant) corresponding to the changes introduced in this PR.
• I've made sure all auto checks have passed.

[sonichi]

A significant amount of effort is devoted to handling the python code which seems to be suitable for executing in a Jupyter Notebook environment. Maybe it'll be cleaner to consider implementing an executor based on ipykernel in future.

The current implementation is better than before. To keep this agent in the core library, the following is required:

1. test code and code coverage
2. be aware of the maintenance requirement. For example, if we make changes to UserProxyAgent, it may break the MathUserProxyAgent. It could happen that we'll need to temporarily disable it until you make it functional again. You'll also be required to review PRs more timely.

The benefit of including it in the core library is to keep it up-to-date and gets improvement over time (including contributions from follow-up research), until it gets actual usage and makes real impact. The associated cost will be the maintenance effort. Please take these into account and decide whether you'd like to keep it in or out of the core library.

[qingyun-wu]

typo in the name of the notebook: "autogent_agent_MathChat.ipynb" should be "autogen_agent_MathChat.ipynb"

[sonichi]

We can disable macos py3.7 test per #1079

[sonichi]

Is it ready to run the OpenAI test?

[sonichi]

Since py3.7 is removed from test, please remove py3.7 from this badge in README.md: