`ClientBootstrap.connect` that returns a NIOAsyncChannel hits precondition if connect fails about swift-nio

[FranzBusch]

Thanks for filing this. We are aware of this problem and @tayloraswift has also reported this in the forums. The problem here is that we tried to model the NIOAsyncChannel has having ownership of the underlying channel so that users have a safe way to interact with it while still making sure the underlying resource (socket,fds,etc.) are cleaned up. However, our current APIs are not enough here since as you reported in the issue there are timing gaps where a NIOAsyncChannel is created and then dropped resulting in our precondition being triggered. There are actually many more of those timing gaps and we have to address them across all the APIs.

My current idea here is that we have to create new bootstrap APIs that provide scope access to the NIOAsyncChannel directly. Those APIs can make sure that the channel is closed after the scope ends. This will allow us to drop the preconditions again.