[Feature Request]: Support for function calling and images in MessageHistoryLimiter

[pamelafox]

Is your feature request related to a problem? Please describe.

We are thinking of using AutoGen for our multi-turn chat application, but we need message limiting based on token count, to avoid going over the context window. I currently use a package that I maintain for that, at https://github.com/pamelafox/openai-messages-token-helper
My package accounts for messages from any GPT model, function calling, and image requests with vision models. I don't believe MessageHistoryLimiter handles function calls and image requests, from what i saw of the code.

Describe the solution you'd like

You could bring in the logic from my package into MessageHistoryLimiter. My logic isn't perfect, since it's based on devs reverse-engineering the OpenAI token construction, but it gets within 4 tokens in all my tests. My repo also includes unit tests and live tests.

Additional context

No response