[asimawdah]

Why

Streaming procedure handlers may need to report errors from callbacks or background tasks that execute after the handler has returned control to the event loop. These errors cannot be handled through the handler's normal error flow, leaving no explicit way to propagate them to the client.

What changed

• Add ctx.uncaught(error) to the procedure handler context to report uncaught server errors and notify the client.
• Reuse the existing UNCAUGHT_ERROR error code and convert the error into a readable message.
• Document how to report errors from callbacks and background tasks.
• Add a regression test verifying that the client receives the expected error and the procedure's AbortSignal is aborted.
• Exclude uncaught from MiddlewareContext.