Force kill unicorn if it still is running

In some situations (e.g. if unicorn isn't listening properly for signals), it's possible the TERM signal isn't enough to kill the unicorn worker. We ran into issues in 10.6.0 where the jemalloc upgrade broke the signal handler, but the script didn't notice the unicorn master wasn't being terminated properly. Modify the wrapper script to wait up to 5 seconds before issuing a kill -9 to the unicorn master.

Closes #3356 (closed)