[Feature Request]: Use Pydantic models for llm_configs, config_list and code_execution_config

[ekzhu]

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

Right now, config_list, llm_configs and code_execution_config are dictionaries, and the schema of these dictionaries are enforced through parser functions.

Describe the solution you'd like

A better way to handle this would be to use Pydantic models so parsing, serializing and data validations are handled by Pydantic. This also address many issues related to invalid configs.

Additional context

Pydantic

Example of Pydantic models already used in our repo: https://github.com/microsoft/autogen/blob/main/autogen/function_utils.py

Existing issues to address data validation:
#1082
#1026
#1074
#1069

[afourney]

Arguably I would add code_execution_config to that as well.