OpenAI Python Agents v0.22.0 Enhances Runtime Hardening and Configuration Validation

Version 0.22.0 of the OpenAI Python Agents SDK introduces substantial runtime hardening and tightens provider configuration contracts. Key changes include redacting sensitive tool outputs from replay state, isolating usage accounting, and expanding agent graph visualizations. Developers combining an explicit `openai_client` with `organization` or `project` must now move these values to the `AsyncOpenAI` client. The release also rejects conflicting provider options and raises `ModelBehaviorError` for failed or incomplete non-streaming responses, improving application reliability and security.