ChatCompletionClient to support request caching

[ekzhu]

Support client-side caching for any ChatCompletionClient type.

Simplest way to do it is to create a ChatCompletionCache type that implements the ChatCompletionClient protocol but wraps an existing client.

Example how this may work:

from autogen_ext.stores.diskcache import DiskCacheStore
from autogen_ext.models.cache import ChatCompletionCache
from autogen_ext.models.openai import OpenAIChatCompletionClient

# Cached client.
cached_client = ChatCompletionCache(OpenAIChatCompletionClient(model="gpt-4o"), store=DiskCacheStore())