Add control-rpc-stop-timeout flag for control socket

[AmitPr]

When a container shuts down with clients connected to the control socket, the runsc process will hang for 15 seconds due to the nonconfigurable shutdown timeout in the urpc server. This PR adds a -control-rpc-stop-timeout flag that allows us to adjust the shutdown timeout.

Context is that we have a resource monitor that creates a persistent connection to the runsc control socket. This connection lasts for the lifetime of the container, we currently send requests to fetch memory usage (approximating runsc usage -full). We can't run runsc usage -full because we manage hundreds of containers per host and runsc usage -full seems to degrade our hosts under load.