[FEATURE] Add support for HF Nvidia NIM API on InferenceEndpointsLLM

[plaguss]

Is your feature request related to a problem? Please describe.
Add support for serverless Nvidia NIM API.

Describe the solution you'd like
As suggested, it will require the following:

The new NIM API requires a specific base_url to be passed:

client = InferenceClient(
    base_url="https://huggingface.co/api/integrations/dgx/v1",
    api_key="MY_FINEGRAINED_TOKEN"
) 

And then it requires a model id to be passed in the model argument of chat_completions:

output = client.chat.completions.create(
    model="meta-llama/Meta-Llama-3-8B-Instruct",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Count to 10"},
    ],
    stream=True,
    max_tokens=1024,
)

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
PR for reference: huggingface/huggingface_hub#2482

[MoritzLaurer]

Thanks! note that the api is paid and it only works if you use a fine-grained token from an enterprise hub org (see details here)