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.