Making it easier to build human-in-the-loop agents with interrupt

LangGraph, an agent framework, was designed with human-in-the-loop workflows in mind. To support these workflows, LangGraph has a persistence layer that allows for checkpointing and resuming of graph execution. A new feature called `interrupt` has been added to facilitate human-in-the-loop interactions in production settings, allowing developers to pause execution, mark the thread as interrupted, and pass user input into the persistence layer. This enables various common human-in-the-loop workflows such as reviewing and approving actions, editing the state of the graph, and conducting multi-turn conversations between agents and humans. The `interrupt` feature aims to make it easier for developers to build agent frameworks that support human collaboration, with plans to release more end-to-end projects demonstrating its capabilities soon.