feat(ci): Add basic integration tests for reasoning using vllm

[s-akhtar-baig]

What does this PR do?

GitHub pull request #3793 added reasoning_content field to the streaming events, but no tests exist to validate the reasoning functionality of Responses API in Llama Stack. OpenAI does not return reasoning tokens, therefore, this PR uses vllm setup to test the basic functionality.

Test Plan

• Added an integration test for basic reasoning streaming events
• Created recordings with --suite reasoning and --setup vllm:

$ uv run ./scripts/integration-tests.sh --stack-config server:ci-tests  --inference-mode record-if-missing --setup vllm --suite reasoning --pattern 'not( builtin_tool or safety_with_image or code_interpreter )'

tests/integration/responses/test_reasoning.py::test_reasoning_basic_streaming[txt=vllm/Qwen/Qwen3-0.6B] 
instantiating llama_stack_client
Port 8321 is already in use, assuming server is already running...
llama_stack_client instantiated in 0.036s
PASSED                                                                                               [100%]

=============================================================================================== slowest 10 durations ===============================================================================================
0.16s setup    tests/integration/responses/test_reasoning.py::test_reasoning_basic_streaming[txt=vllm/Qwen/Qwen3-0.6B]
0.12s call     tests/integration/responses/test_reasoning.py::test_reasoning_basic_streaming[txt=vllm/Qwen/Qwen3-0.6B]

(1 durations < 0.005s hidden.  Use -vv to show these durations.)
================================================================================================ 1 passed in 0.29s =================================================================================================
+ exit_code=0
+ set +x
✅ All tests completed successfully

=== Integration Tests Complete ===