Catch signals and allow managed tearDown and reloads

[theopolis]

Catching signals is helpful! This allows us to gracefully call tearDown on all plugins but most importantly the EventPublisherPlugin types. The soon-to-be-merged audit publisher for Linux (#1463) makes system configuration changes during setUp that SHOULD be removed during tearDown. The HUP signal could be forwarded to the active ConfigPlugin and cause a plugin-optional config reload.

When a worker/watcher is used the parent should forward handled signals to the worker process.

[theopolis]

Signal handling was implemented in #1494. There are some locations where event publishers do not tear down within a reasonable amount of time. To extend the scope of this issue a bit

• Test that Linux and Darwin-based publishers tearDown consistently within the alarm time of 10 seconds.
• Throw the original signal instead of alarm if the tearDown grace period expires.
• Modify the Linux init script and osqueryctl to send non-kill commands, which allow graceful shutdown.