Fix inference search

[Wauplin]

EDIT: converting to draft for now. Turns out

• https://huggingface.co/api/models/stabilityai/stable-diffusion-xl-base-1.0?expand[]=inferenceProviderMapping
• https://huggingface.co/api/models?search=stable-diffusion-xl-base-1.0&inference=warm&expand[]=inferenceProviderMapping

do not return in the same format.

---

Fix #2963.

With this PR:

• inference parameter accept "warm" or None when listing models
• possible to filter models by inference_provider (one, multiple or "all")
• return value is now a list, not a dict anymore (server-side breaking change)

from huggingface_hub import list_models

for model in list_models(inference_provider=["cerebras"], expand=["inferenceProviderMapping"], limit=1):
    print(model.id)
    print(model.inference_provider_mapping)
    break

=>

meta-llama/Llama-4-Scout-17B-16E-Instruct
[
    InferenceProviderMapping(provider='novita', provider_id='meta-llama/llama-4-scout-17b-16e-instruct', status='live', task='conversational'),
    InferenceProviderMapping(provider='fireworks-ai', provider_id='accounts/fireworks/models/llama4-scout-instruct-basic', status='live', task='conversational'),
    InferenceProviderMapping(provider='together', provider_id='meta-llama/Llama-4-Scout-17B-16E-Instruct', status='live', task='conversational'),
    InferenceProviderMapping(provider='sambanova', provider_id='Llama-4-Scout-17B-16E-Instruct', status='live', task='conversational'),
    InferenceProviderMapping(provider='cerebras', provider_id='llama-4-scout-17b-16e-instruct', status='live', task='conversational')
]

@hanouticelina will be useful to merge into #3011

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.