add notify_closing

[graingert]

Changes

• add anyio.notify_closing
• adjust wait_readable/wait_writable to checkpoint after registration and before waiting on the event.

anyio-zmq needs this feature to handle ZmqSocket.close calls

we expose the trio implemenation which is simple enough, and emulate the feature on asyncio. If someone is using add_reader themselves on an FD then you can't use it with anyio wait_readable/wait_writiable/notify_closing

Checklist

If this is a user-facing code change, like a bugfix or a new feature, please ensure that
you've fulfilled the following conditions (where applicable):

• You've added tests (in tests/) added which would fail without your patch
• You've updated the documentation (in docs/, in case of behavior changes or new
  features)
• You've added a new changelog entry (in docs/versionhistory.rst).

If this is a trivial change, like a typo fix or a code reformatting, then you can ignore
these instructions.

Updating the changelog

If there are no entries after the last release, use **UNRELEASED** as the version.
If, say, your patch fixes issue #123, the entry should look like this:

- Fix big bad boo-boo in task groups
  (`#123 <https://github.com/agronholm/anyio/issues/123>`_; PR by @yourgithubaccount)

If there's no issue linked, just link to your pull request instead by updating the
changelog after you've created the PR.

[agronholm]

So, two questions came to mind:

1. Is there a reason we're not just popping the fd from the dict when we have a result?
2. Is there a reason why you're not simply setting the ClosedResourceError as the future's exception?

[graingert]

So, two questions came to mind:

1. Is there a reason we're not just popping the fd from the dict when we have a result?

that seems nicer, and allows new readers/writers slightly sooner. I've pushed that - it's a bit more complex though

2. Is there a reason why you're not simply setting the `ClosedResourceError` as the future's exception?

using a future for handling exceptions results in a reference cycle attached to a traceback and I'd like to avoid this. Also the traceback is harder to debug