fix served_model_name

[yangli5t]

Why ?

to make served_model_name working properly. with this arg, it allows pass in a customized model name when query the model.
arguments doc: https://docs.vllm.ai/en/v0.8.3/serving/openai_compatible_server.html
implementation was referred to https://github.com/vllm-project/vllm/blob/73a484caa1ad320d6e695f098c25c479a71e6774/vllm/entrypoints/openai/api_server.py#L1090-L1092

How ?

simply add "served_model_name" in the command, e.g. python -m matrix deploy_applications --applications '[{"model_name": "meta-llama/Llama-3.1-8B-Instruct", "model_size": "8B", "name": "8B", "min_replica": 1, "max_replica": 10, "enable_tools": "true", "served_model_name": "8B-matrix"}]'
then in query_llm.make_request, model="8B-matrix"

Test plan

end-to-end test

response = client.chat.completions.create(
        model='8B-matrix',
        messages=messages,
        temperature=0.6,
        top_p=0.95,
)