Feature Request] Support Webhook/Callback for long-running code execution results

Feature Description

Support for asynchronous code execution with a callback/webhook mechanism.

Use Case

I'm building an agent that performs long-running tasks (e.g., 10+ minutes of data analysis). Currently, I need to keep the SDK connection open to receive results via on_result or on_stdout. For these long tasks, a persistent connection is unreliable.

Proposed API

It would be great if sandbox.run_code or a similar method could accept a callback configuration:

# Conceptual example
sandbox.run_code(
    code="very_long_task()",
    callback_url="[https://my-api.com/v1/e2b-webhook](https://my-api.com/v1/e2b-webhook)",
    metadata={"task_id": "123"}
)

Prüfe zuerst die bestehenden APIs sandbox.run_code, on_result und on_stdout. Definiere, wie callback_url und metadata akzeptiert werden sollen und wie Ergebnisse lang laufender Ausführungen über den Webhook übermittelt werden, und verifiziere anschließend, dass Clients die SDK-Verbindung nicht mehr offen halten müssen.