ChatCompletionCache should support component config

[jackgerrits]

The new cache should support declarative configuration

[victordibia]

Small side comment ... @jackgerrits
Would it be possible to have a simpler interface for enabling cache in the chatcompletion client?

cached_model_client = OpenAIChatCompletionClient(model="gpt-4o", cache=True)

This way ChatCompletionClient caching can be reflected in the config for the client.

[jackgerrits]

This means that each model client need to implement it. The point of the current design is separation of concerns and generalizability.

This way ChatCompletionClient caching can be reflected in the config for the client.

This is the case currently - itll be a model client's config nested with the config of the cache client.

If we allow a default cache store, then the current usage can be as simple as:

cached_model_client = ChatCompletionCache(OpenAIChatCompletionClient(model="gpt-4o"))