Update openai test

[XG-xin]

Motivation

Move reasoning tokens from metadata to metrics in nodejs openai tests.
Need to configure to skip some old dd-trace versions.

Changes

Workflow

1. ⚠️ Create your PR as draft ⚠️
2. Work on you PR until the CI passes
3. Mark it as ready for review
   • Test logic is modified? -> Get a review from RFC owner.
   • Framework is modified, or non obvious usage of it -> get a review from R&P team

🚀 Once your PR is reviewed and the CI green, you can merge it!

🛟 #apm-shared-testing 🛟

Reviewer checklist

• If PR title starts with [<language>], double-check that only <language> is impacted by the change
• No system-tests internal is modified. Otherwise, I have the approval from R&P team
• A docker base image is modified?
  • the relevant build-XXX-image label is present
• A scenario is added (or removed)?
  • Get a review from R&P team

[github-actions]

CODEOWNERS have been resolved as:

tests/integration_frameworks/llm/openai/test_openai_llmobs.py           @DataDog/ml-observability

[sabrenner]

I think what we can do to best avoid issues with updating the repos is

1. Split out the Responses tests into their own test class in test_openai_llmobs.py, ie

@features.llm_observability_openai_llm_interactions
@scenarios.integration_frameworks
class TestOpenAiResponses(BaseOpenaiTest):

then in each of the nodejs.yml and python.yml manifests, we can mark those test classes as irrelevant

test_openai_llmobs.py:
  TestOpenAiEmbeddingInteractions: *ref_5_80_0
  TestOpenAiLlmInteractions: *ref_5_80_0
  TestOpenAiPromptTracking: missing_feature
  TestOpenAiResponses: irrelevant

and similarly for python.yml. We can update these once the features land in a given release.

This will resolve the responses tests. For the other tests where we assert reasoning tokens are present but 0, we can do

assert_llmobs_span_event(
  ...
  metrics=mock.ANY
)

# assert input, output, total, and maybe cached tokens separately
assert llm_span_event["metrics"]["input_tokens"] = ...

to make the tests version-independent and more resilient to future metrics, as we only want to assert specific tokens in those tests. lmk if this makes/doesn't make sense!