[CaptainAgent] Add executor injected with tools.

[LeoLjl]

Current tools in ag2 can only be called via function calling/tool calling. This relies on the capability of the backend LLM. Another way of orchestrating tools is to use python coding to call and orchestrate the results. Introducing LocalExecutorWithTools, a customized executor that can call the functions underling the ag2 tools. The executor takes in a list of ag2 tools when instantiated, and extract the func field within them. An example usage is:

api_wrapper = WikipediaAPIWrapper(top_k_results=1, doc_content_chars_max=3000)
langchain_tool = WikipediaQueryRun(api_wrapper=api_wrapper)
interop = Interoperability()
ag2_tool = interop.convert_tool(tool=langchain_tool, type="langchain")

# `ag2_tool.name` is wikipedia
local_executor = LocalExecutorWithTools(tools=[ag2_tool], work_dir="./")

code = """
result = wikipedia(tool_input={"query":"Christmas"})
print(result)
"""
result = local_executor.execute_code_blocks(
    code_blocks=[
        CodeBlock(language="python", code=code),
    ]
)

On the side of agents, they just need to call the function with no declarations needed and the results will be provided. The provides a more LLM-friendly way of orchestrating tool usage.

This PR is the preliminary of integrating cross platform tool into CaptainAgent. This is because the agents in groupchat assembled by CaptainAgent use tools as freeform coding. In order to adapt the tool-calling based ag2 tools into CaptainAgent, an executor injected with the namspace of the tool names is needed.

Why are these changes needed?

Related issue number

#286

Roadmap

The next PR will add the full content of the integration code and the notebook demonstrating the usage.

Checks

• I've included any doc changes needed for https://docs.ag2.ai/. See https://docs.ag2.ai/docs/contributor-guide/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.

[skzhang1]

Looks good to me.