[3/6] Run Python SDK tests in CI

[aibrahim-oai]

Stack

• [1/6] [1/6] Pin Python SDK runtime dependency #21891 Pin Python SDK runtime dependency
• [2/6] [2/6] Generate Python SDK types from pinned runtime #21893 Generate Python SDK types from pinned runtime
• [3/6] [3/6] Run Python SDK tests in CI #21895 Run Python SDK tests in CI
• [4/6] [4/6] Define Python SDK public API surface #21896 Define Python SDK public API surface
• [5/6] [5/6] Rename Python SDK package to openai-codex #21905 Rename Python SDK package to openai-codex
• [6/6] [6/6] Default Python SDK approval policy to never #21910 Default Python SDK approval policy to never

This PR is the third change in the stack, on top of #21893.

Why

The Python SDK tests need to run online so the dependency pin, pinned-runtime type generation, async coverage, and interleaving coverage stay enforced after review.

What changed

• Add a separate python-sdk job to .github/workflows/sdk.yml.
• Run the job in python:3.12-alpine so the pinned musl runtime wheel installs normally in CI.
• Run the Python SDK pytest suite in parallel with the existing SDK job.

Verification

• python-sdk passed in CI on this PR.
• sdks passed in CI on this PR.

[aibrahim-oai]

This change is part of the following stack:

• Publish Python runtime wheels on release #21812
  • [1/6] Pin Python SDK runtime dependency #21891
    • [2/6] Generate Python SDK types from pinned runtime #21893
      • [3/6] Run Python SDK tests in CI #21895 ◀

_{Change managed by git-spice.}