update vllm from 0.11.2 to 0.12 for NVIDIA configs

[functionstackx]

vllm-project/vllm@releases/v0.11.2...releases/v0.12.0

https://hub.docker.com/layers/vllm/vllm-openai/v0.12.0/images/sha256-f2309d913a07da49ea20b2a694703f4cfcb5ad8e7437ec0f26145479ac01e002

[functionstackx]

seems like same recipe as 0.11.2 https://github.com/vllm-project/recipes/blob/main/OpenAI/GPT-OSS_Blackwell.yaml

[functionstackx]

done in https://github.com/InferenceMAX/InferenceMAX/pull/327