Built an Execution Control Foundation for AI Conversation

Built an Application Execution Protocol foundation for controlling the execution flow of AI conversation processing, including retry, loop, and stop behavior through a shared mechanism.

Execution patterns such as retrying Content Review, re-running the full conversation based on a Runtime Review result, and safely stopping when MaxAttempts is reached are now handled as common execution control, separated from AI conversation-specific review decisions.

This makes it possible to define clearly in code how review decisions are reflected in execution and where processing should resume. Compatibility with human conversation sending and the existing Conversation Runtime is maintained.