Multimodal conversable agents

[BeibinLi]

### Tasks
- [x] Create a multi-modal agent and test
- [x] Test with GPT4-V API
- [ ] https://github.com/microsoft/autogen/issues/547

[sonichi]

built by community:
vision: https://www.youtube.com/watch?v=zo2ft4Qje1Y
speech: https://www.youtube.com/watch?v=JgVb8A6OJwM

[gagb]

@BeibinLi this issue was marked done and closed but shouldn't that be after multimodal agents become a core part of repo? If this issue was specific to the notebook example, should we have an open issue on the roadmap for multimodal conversable agents?