feat: add oci genai service as chat inference provider

[dkennetzoracle]

What does this PR do?

Adds OCI GenAI PaaS models for openai chat completion endpoints.

Test Plan

In an OCI tenancy with access to GenAI PaaS, perform the following steps:

1. Ensure you have IAM policies in place to use service (check docs included in this PR)
2. For local development, setup OCI cli and configure the CLI with your region, tenancy, and auth here
3. Once configured, go through llama-stack setup and run llama-stack (uses config based auth) like:

OCI_AUTH_TYPE=config_file OCI_CLI_PROFILE=CHICAGO OCI_REGION=us-chicago-1 OCI_COMPARTMENT_OCID=ocid1.compartment.oc1..aaaaaaaa5...5a llama stack run oci

4. Hit the models endpoint to list models after server is running:

curl http://localhost:8321/v1/models | jq
...
{
      "identifier": "meta.llama-4-scout-17b-16e-instruct",
      "provider_resource_id": "ocid1.generativeaimodel.oc1.us-chicago-1.am...q",
      "provider_id": "oci",
      "type": "model",
      "metadata": {
        "display_name": "meta.llama-4-scout-17b-16e-instruct",
        "capabilities": [
          "CHAT"
        ],
        "oci_model_id": "ocid1.generativeaimodel.oc1.us-chicago-1.a...q"
      },
      "model_type": "llm"
},
   ...

5. Use the "display_name" field to use the model in a /chat/completions request:

# Streaming result
curl -X POST http://localhost:8321/v1/chat/completions   -H "Content-Type: application/json"   -d '{
        "model": "meta.llama-4-scout-17b-16e-instruct",
       "stream": true,
       "temperature": 0.9,
      "messages": [
         {
           "role": "system",
           "content": "You are a funny comedian. You can be crass."
         },
          {
           "role": "user",
          "content": "Tell me a funny joke about programming."
         }
       ]
}'

# Non-streaming result
curl -X POST http://localhost:8321/v1/chat/completions   -H "Content-Type: application/json"   -d '{
        "model": "meta.llama-4-scout-17b-16e-instruct",
       "stream": false,
       "temperature": 0.9,
      "messages": [
         {
           "role": "system",
           "content": "You are a funny comedian. You can be crass."
         },
          {
           "role": "user",
          "content": "Tell me a funny joke about programming."
         }
       ]
}'

6. Try out other models from the /models endpoint.

[ashwinb]

@github-actions run precommit

[github-actions]

⏳ Running pre-commit hooks on PR #3876...

[github-actions]

✅ Pre-commit hooks completed successfully!

🔧 Changes have been committed and pushed to the PR branch.

[dkennetzoracle]

Removing docs additions at request of @raghotham

[ashwinb]

cc @mattf for a review since this touches the inference system

[dkennetzoracle]

Any updates here?

[leseb]

Suggested change

	class OCIConfig(BaseModel):
	class OCIConfig(RemoteInferenceProviderConfig):

[leseb]

Why comment out? Are all of them unsupported?

[leseb]

We have OpenAIMixin Class that exposes a lot of knobs, cna you look at it and see if you can use it? Instead of using "custom" code to do the completion requests?

[mattf]

@dkennetzoracle -

1. does oci provide an openai compatible endpoint?
2. please include output of the inference tests against the remote::oci provider

[ashwinb]

I think it would be much preferable if we can work against an OpenAI compatible endpoint. Otherwise, at the very least we need a set of recorded tests against the provider. But before recordings, let's make sure at least the tests pass "live". Here's a command to run (roughly):

pytest -sv tests/integration/inference/  \
   --stack-config <your_distro> \
   --text-model <oci/...>   \
   --embedding-model sentence-transformers/nomic-ai/... \
   --inference-mode live

[dkennetzoracle]

@leseb @ashwinb @mattf thanks for reviewing. If it would be strongly preferable for me to use an OpenAI compatible endpoint, I can make those changes. I'll refactor and re-request when this is done.

Sorry also, I started the PR a few weeks ago before a conference, and when I got back inference providers had changed significantly, although it seems like for the better. I'll align on the changes and re-request!