Fix flakiness in testWithConfiguredStreamSocket

[Lukasa]

Motivation:

This test occasionally fails in our ongoing continuous CI solution. Flaky tests are bad, and we should aim not to have them, so let's fix it.

My diagnosis of the most likely race here (which leads to alreadyClosed) being thrown somewhere from this code is that the client channel got closed unexpectedly. This would happen if an unusual series of events occurs:

1. The client connection succeeds and the channel starts up
2. The server close then goes through and the server channel is closed before it actually accepts the connection.
3. This causes the client connection to be reset.
4. The client channel observes this and handles that close.
5. We then close the client from the outside.

Modifications:

Add a ConditionLock we can use to wait for a connection to be established.
Make the use of this ConditionLock resilient to weird test behaviour by always using timeouts for blocking operations.

Result:

Less flaky tests.