Cerebras Integration

[henrytwo]

Why are these changes needed?

Add integration for Cerebras, which provides super low-latency, high speed LLM inference. Currently Llama 3.1-8B/70B are supported.

Tool calling examples have been provided in this PR, which also function with streaming. Additionally, token cost calculations have been implemented.

Checks

• I've included any doc changes needed for https://microsoft.github.io/autogen/. See https://microsoft.github.io/autogen/docs/Contribute#documentation to build and test documentation locally.
• I've added tests (if relevant) corresponding to the changes introduced in this PR.
• I've made sure all auto checks have passed.

[Hk669]

im so amazed with the inference speed. Great work cerebras, @henrytwo . Ive made few suggestions and also we dont need the autobuild notebook formatting in this. We can have a seperate PR.

cc @marklysze

[Hk669]

i've tested with the chess tool call example, a few general examples and marketing campaign use cases. the speed of the inference is super impressive. thanks @henrytwo.

Looks good to me.