When Hazelcast is started on a fixed port and that port is already in use, it does not fail to start which makes the problem hard to detect. A symptom of the problem is that clients will not be able to connect and will either retry indefinitely or will timeout depending on their configuration.
This PR updates the Hazelcast client tests to start the Hazelcast instance on an ephemeral port and to customize the client configuration with the instance's address before use. This should allow the client tests to work reliably in an environment where Hazelcast's default port is already in use. The hope is that it will fix the PR build that runs on GitHub Actions where HazelcastAutoConfigurationClientTests currently fails.