Token counter to Agentes

# Description:
We propose implementing a token counter functionality within Agentes. This feature aims to collect all the prompt and output tokens of the large language model (LLM) and include them in the Agent executor chain result upon calling the Call function.

This enhancement would provide valuable insights into the token usage patterns of the LLM during execution, which can be beneficial for various purposes such as analyzing model behavior, optimizing resource utilization, and debugging.

Look for the Agent executor chain and the Call function in the codebase to understand how prompts and outputs are currently handled. Identify where LLM interactions occur to instrument token counting. The result structure needs to be extended to include token counts. Check existing tests for the executor to see how to validate the new functionality.

由索引模型根據 Issue 內容生成。