fix(core): Stop container when signals captured

[lamcw]

The current atexit hook that stops (and optionally remove) containers in the Docker daemon would not work because of limitations of the atexit module: the handlers are not executed if the Python interpreter is

killed by a signal not handled by Python, when a Python fatal internal error is detected, or when os._exit() is called.
https://docs.python.org/3/library/atexit.html

To handle shutdowns of containers completely (for e.g. when the tests are terminated via SIGINT / SIGTERM), we'd need to register signal handlers for them as well.

This kind of acts as an interim solution while #314 is being implemented.

[lamcw]

A more advanced version of signal.signal that does not overwrite existing signal handler registered to signal.

[santi]

This issue will be resolved when #314 lands, which removes the atexit hook completely and removes the dependency on GC to cleanup danglings containers

[alexanderankin]

going to close since reaper v1 has landed, and even reaper v2 (only difference being auto_remove=True (aka docker run --rm) so the instances dont hang out and haunt your system 👻)

if any other issues with stopping containers or anything else, just give a shout! @lamcw