feat(openai): stream Responses across IRR agentic rounds

[leseb]

Summary

Adds terminal and multi-round agentic Responses SSE streaming through IRR using praxis-proxy/praxis#988 at a198962d, preserving one logical response across MCP and web-search rounds.
Includes bounded lifecycle and error handling, persistence integration, example configs, generated docs, synthetic inference fixtures, and live vLLM terminal-streaming coverage.

Related issue

Closes #706

Validation

• Unit tests: cargo test -p praxis-ai-apis openai::responses::agentic_loop::tests and cargo test -p praxis-ai-apis openai::responses::stream_events::tests
• Integration tests: cargo test -p praxis-tests-integration --test suite examples::openai_agentic_loop -- --nocapture and cargo test -p praxis-tests-integration --test suite examples::irr_terminal_streaming -- --nocapture
• Inference fixtures: make test-inference-fixtures
• Live backend: uv run tests/integration/sdk/openai/test_openai_responses_vllm.py -s -vv -k test_streaming_through_irr
• make build
• make lint

Checklist

• I reviewed every changed line and can explain the change.
• New capabilities include an example config and functional example test.
• User-facing behavior and generated documentation are updated.
• Performance-sensitive changes include appropriate benchmark or load-test evidence.
• Commits are signed and include a Signed-off-by trailer.

Breaking changes

None.