feat(engine)!: execute loop and iteration in engine frames

[laipz8200]

Important

1. Make sure you have read our contribution guidelines
2. Search existing issues and pull requests to confirm this change is not a duplicate
3. Open or identify the issue this pull request resolves or advances
4. Use a Conventional Commits title for this pull request, and mark breaking changes with !
5. Remember that the pull request title will become the squash merge commit message
6. If CLA Assistant prompts you, sign CLA.md in the pull request conversation

Related Issue

close #198

Summary

• Executes loop and iteration containers inside GraphEngine frames instead of nested engines.
• Adds frame-aware ready tasks, event routing, and child frame materialization for container bodies.
• Makes DSL loading understand loop and iteration container node types.

Breaking Changes

• LoopNode and IterationNode no longer execute child graphs from their own _run() methods. Their bodies are now interpreted by GraphEngine through frame-aware execution.
• Runtime state no longer exposes the child-engine execution path used by the old loop and iteration implementation.
• Engine-internal task routing now carries frame_id with node_id. Custom ready queue, dispatcher, worker, or event-handler integrations must preserve the frame when scheduling or handling node work.
• Container child graphs are materialized by rebinding the graph node factory to a frame-local GraphRuntimeState. Custom node factories used with loop or iteration graphs must provide with_runtime_state(graph_runtime_state).
• NodeExecution.execution_id is now required when the domain model is created, and NodeExecution.mark_started() no longer accepts an execution id.
• Nodes no longer create or restore execution ids through ensure_execution_id; the method has been removed. The worker binds the frame-scoped execution id before calling node.run().

Downstream Adaptation

• Run complete graphs through GraphEngine.run() or graphon.dsl.loads(...).run() instead of calling LoopNode.run() or IterationNode.run() directly.
• Treat emitted graph events as the public execution record. Child node events inside containers still surface through the engine event stream and carry in_loop_id or in_iteration_id.
• If you implement a custom ready queue, store and return ReadyTask(frame_id, node_id) values. Use root for normal top-level work.
• If you implement a custom node factory and need loop or iteration support, add a with_runtime_state() method that returns an equivalent factory bound to the supplied runtime state.
• If you instantiate NodeExecution directly, provide execution_id up front. Prefer GraphExecution.get_or_create_node_execution() when possible.
• Replace node_execution.mark_started(execution_id) with node_execution.mark_started().
• Replace direct node.ensure_execution_id() usage with worker-style binding: get the frame-scoped node execution, call node.bind_execution_id(node_execution.execution_id), then call node.run().

Checklist

• This pull request links the issue it resolves or advances
• This pull request title follows Conventional Commits, and any breaking change is marked with !
• If CLA Assistant prompted me, I signed CLA.md in the pull request conversation