Switch container steps to use an active listener. By default, this step runs the equivalent of kubectl exec to run nested sh steps inside the designated container. While that works, it is inefficient as it forces the Kubernetes API server to handle traffic for each such step. When active container mode is enabled, any non-agent container whose command is sleep (or cat) is made to instead run a listening script which awaits requests by the agent container. This is more efficient and scalable.

Note: this mode is currently supported only on Linux. Windows containers will use the original implementation.