Add aclose to AsyncNetworkBackend interface

[johnjhughes]

Summary

• add an aclose method to the AsyncNetworkBackend interface to make the async backend contract explicit for type checkers

Testing

• not run (interface-only change)

[Copilot]

Pull request overview

This PR adds an aclose method to the AsyncNetworkBackend interface to make the async backend contract explicit for type checkers.

Changes:

• Added an aclose method to the AsyncNetworkBackend class that raises NotImplementedError

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Adding this method to the AsyncNetworkBackend interface creates a breaking change for type checkers. None of the existing implementations (AnyIOBackend, TrioBackend, AutoBackend, AsyncMockBackend) have implemented this method, which means:

1. Type checkers will report that these implementations are missing the required aclose method
2. Test backends in tests/test_cancellations.py (SlowWriteBackend, SlowReadBackend) will also fail type checking

Additionally, there is no usage of backend.aclose() anywhere in the codebase (e.g., connection pools do not call aclose on their backends), which suggests this method may not be needed yet.

If the goal is to make backends closeable for type checking purposes, all implementations need to be updated in the same PR to include either a working implementation or a no-op pass statement. The AutoBackend implementation in particular would need to delegate to self._backend.aclose() if that backend exists.

Suggested change

	async def aclose(self) -> None:
	raise NotImplementedError() # pragma: nocover

[Copilot]

This change creates an asymmetry in the API design. The synchronous NetworkBackend class (line 36) does not have a close method, but now AsyncNetworkBackend has an aclose method. This inconsistency may be confusing for users and library maintainers.

Consider whether:

1. NetworkBackend should also have a close method for consistency
2. Or if neither backend type needs lifecycle management methods since they are typically stateless factories

Note that both NetworkStream (line 21) and AsyncNetworkStream (line 66) have close/aclose methods respectively, so the current pattern is that streams (connections) need cleanup but backends (connection factories) do not.