Introduce a watchdog thread to handle termination gracefully

[abbysmal]

This addresses issue #2 as well.

The source of the segfault was that, on ctrl-c, we would often find ourselves in situations where we can close the Midi.device twice. (once in the signal handler, once at the termination of the main thread, that is also calling Midi.shutdown.)

Another source of segfault was having the sequencer thread in stat_engine write to a device that was closed by said signal handler.

I also noticed that the stat_engine would not check the child_alive function to see if the instrumented program was still running or not, leading to cardio-crumble running indefinitely trying to poll events that would never show up.

I decided to implement a solution to all of these problems by adding a watchdog thread to be run alongside any threads the engine is using.

The idea being it is that it would introduce a Watchdog.terminate atomic variable, that each thread should check from time to time. If Watchdog.terminate is true, then said threads should be exiting gracefully.

The watchdog domain itself is just checking that child_alive is false (meaning the instrumented program has stopped).

If it detects it is the case, it will then set terminate to true, and exit.

This may be an overkill solution, but I think it works quite nicely, let me know if you have any nicer idea. :-)

[pitag-ha]

Yeaaaaaaaaaaaaaaaaah! :)