OpenAI Agents SDK v0.18.2: GPT-5.6 Request Control & Managed Multi-Agent Beta

This update adds support for GPT-5.6 request control, allowing developers to directly manage request parameters for GPT-5.6 via the SDK. It also introduces the managed multi-agent Beta, providing infrastructure for building complex multi-agent systems. Fixes include improved real-time callback error handling, replay timing, and sandbox PTY/Docker cleanup ownership; it also fixes response issues in litellm integration for content filter rejections. Refactoring unifies sandbox PTY output collection, rclone setup for E2B/Runloop, and retry error metadata extraction/parsing, enhancing code maintainability. For developers building automated workflows with OpenAI Agents SDK, GPT-5.6 request control directly impacts agent reasoning and output quality. The managed multi-agent Beta reduces complexity in deploying collaborative agents, suitable for scenarios requiring task division. Sandbox and replay fixes also boost stability for real-time interactive applications.