fix(openai): read vllm reasoning deltas

[he-yufeng]

Summary

Fixes #2182.

OpenAI-compatible vLLM streams can emit reasoning text on delta.reasoning instead of the older delta.reasoning_content field. The OpenAI adapter now accepts both fields and still prefers the existing reasoning_content path when present.

Changes

• Read non-empty string reasoning from delta.reasoning_content or delta.reasoning.
• Keep existing text/tool streaming behavior unchanged.
• Add regression coverage for a vLLM-style delta.reasoning stream.

To verify

Run from strands-py:

uv run pytest tests/strands/models/test_openai.py -q --basetemp .tmp/pytest-2182 -p no:cacheprovider
uv run ruff check src/strands/models/openai.py tests/strands/models/test_openai.py
python -m py_compile src/strands/models/openai.py tests/strands/models/test_openai.py
git diff --check

Local result on Windows: 96 passed.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!