Add support for deploying an experiment to an existing ROS network

[finger563]

If the user provides (in the configuration dialog) the ip address and port number of a running roscore, then we should not spawn a roscore and should instead point the newly deployed nodes to it. Then the stop experiment will automatically not find a node running roscore and will handle it fine.

This enables users to break up their experiments into long-running, stable components that can be left on the system and deployments containing component-under-test. It also allows deploying onto an already existing system and further mixing of ROSMOD and non-ROSMOD executables.