Fixed hanging for NApp exception

[Alopalao]

Closes #601

Summary

Starting controller after adding the signals.

Local Tests

Excercise this PR with wrong values in settings.py so mef_eline would failed start.

End-to-End Tests

N/A

[viniarck]

Thanks for the quick fix @Alopalao, it's in the right direction but it's caused other issues. start_shell_async has to run after controller.start() coro, otherwise it'll crash due to server attrs dependency it access, and the shell won't start correctly:

shell_task: <Task finished name='Task-1' coro=<start_shell_async() done, defined at /home/viniarck/repos/kytos/kytos/core/kytosd.py:136> exception=AttributeError("'NoneType' object has no attribute 'server_address'")>

Try moving shell_task after controller.start(), and then make sure to explore all the related cases that you've been fixing and exploring

[Alopalao]

I misunderstood the use of sys.exit(). SystemExit was catching the raise not the signals. Now it runs as expected.

• I also tried to start controller inside create_task() instead of run_until_complete but the exiting was printing the reason of the crash.
• Also tried separating dependency from the shell to obtain adress and port from controller.options instead of the server. But, after exercising the crash, this was leaving any terminal not able to show any input. (Input was not showing as in password masking or input hiding in console)

[viniarck]

Nicely done @Alopalao, really great to have this regression fixed.

Before merging, check out the linter error.

After you fix it, go ahead and merge into master.