feat: Add enable_model_discovery to enable/disable model discovery on startup

[akram]

• Add enable_model_discovery configuration flag to vLLM provider to control model listing behavior
• Implement enable_model_discovery() method across all providers with default implementations in base classes
• Prevent HTTP requests to /v1/models endpoint when enable_model_discovery=false to prevent crash on startup

What does this PR do?

Allows to disable model listing on startup. Useful for models that are declared in vllm but not reachable (not configured or behind authentication wall)
participate in fixing #3151

Test Plan

providers:
  inference:
  - provider_id: ${env.VLLM_URL:+vllm}
    provider_type: remote::vllm
    config:
      url: ${env.VLLM_URL:=}
      max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
      api_token: ${env.VLLM_API_TOKEN:=fake}
      tls_verify: ${env.VLLM_TLS_VERIFY:=true}
      refresh_models: ${env.VLLM_REFRESH_MODELS:=false}
      enable_model_discovery: ${env.ENABLE_MODEL_DISCOVERY:=false}

models:
- metadata:
    display_name: vllm
  model_id: vllm
  provider_id: vllm
  model_type: llm
      

run

LLAMA_STACK_LOGGING="all=debug" VLLM_URL=https://my-vllm-server:8443/v1  MILVUS_DB_PATH=./milvus.db INFERENCE_MODEL=vllm uv run --with llama-stack llama stack build --distro starter --image-type venv --run

you should see:

DEBUG    2025-10-06 17:06:43,560 llama_stack.providers.remote.inference.vllm.vllm:286 inference::vllm: VLLM list_models called, enable_model_discovery=False
DEBUG    2025-10-06 17:06:43,561 llama_stack.providers.remote.inference.vllm.vllm:288 inference::vllm: VLLM list_models returning None due to enable_model_discovery=False```

[akram]

@ashwinb can you PTAL ?

[mattf]

the logic around model management is already complex with multiple knobs.

can you use allowed_models instead?