fix(runtime-sdk): ensure finalize_request is alive inside wait_until call

[ryanking13]

While working on cloudflare/python-workers-examples#103, I noticed that finalizer_task is gc-ed and cleaned up improperly inside wait_until.

This fixes it by wrapping it with create_proxy.

Probably the long term solution is to land #90 and make self.ctx.waitUntil and workers.wait_until work exactly the same, but #90 requires JSPI which isn't available in 0.26.0a2, so it needs more discussion.

[ask-bonk]

create_proxy() transfers ownership to the caller, so this proxy is retained permanently for every streaming response. Destroy it when the background finalizer settles, as the websocket path does.

Suggested change

	task_proxy = create_proxy(finalizer_task)
	finalizer_task = run_in_background(finalize_request())
	task_proxy = create_proxy(finalizer_task)
	finalizer_task.add_done_callback(lambda _: task_proxy.destroy())
	wait_until(task_proxy)

[ask-bonk]

I'm Bonk, and I've done a quick review of your PR.

Ensures streaming ASGI finalizers remain alive across wait_until.

1. P2: The new finalizer proxy is never destroyed, leaking one Pyodide proxy per streaming request. Posted one inline suggestion.

github run

[dom96]

AI comment might have a point, but rest looks good so approving.

[hoodmane]

This is a duplicate of #249. I think it should be fixed in workerd:
cloudflare/workerd#7293

[ryanking13]

This is a duplicate of #249. I think it should be fixed in workerd:
cloudflare/workerd#7293

sounds good. Thanks for fixing that