Implementing Support for Streaming Intermediate Agent Steps

Hello Rust Port Team,

I am currently enhancing the Rust port of the Lagchain library and I'd like to propose the implementation of a feature that is already present in the original Lagchain library: the ability to stream intermediate steps of agents. This feature is invaluable for, Impruving the UX, debugging, monitoring, and scenarios where intermediate data is crucial for understanding the agent's progression through its tasks.

In the original Lagchain library, this functionality facilitates a deeper insight into the agent's operation by providing real-time data at various checkpoints. Integrating this into the Rust port could not only align our feature set more closely with the original library.

Proposed Feature:
Integrate a mechanism to stream intermediate steps of agents in the Rust port, similar to the functionality available in the original Lagchain library.

Look at the original LangChain library's streaming implementation for agents. Examine the Rust port's agent and step execution code to understand the current flow. Identify where intermediate data can be captured and emitted as a stream. A 'done' state would be a working example showing real-time step data during an agent run.

由索引模型根據 Issue 內容生成。