Kill switch: Polish and integrate MVP version

[facuspagnuolo]

Follow up on #516

Based on an offline discussion we had about the next-steps regarding the kill-switch topic, we decided to:

• Drop the kernel level kill-switch idea, at least for the MVP version
• Simplify the whole architecture to support only application level kill-switch using severities
• Support having one issue registry per application
• Integrate with the rest of the aragonOS components
• Investigate how an organization is able to recover

Summing up, the new contracts model looks like:

And the contracts interaction is the following:

[sohkai]

There's a bunch of really small comments in this review, but the main ones I'd like to bring up:

• Making the interface a bit more generic; rather than constraining it to just the kill switch perhaps we can operate on the level of an app being "enabled" or "disabled" by external factors
• Perhaps we can make the check for the KillSwitch a view interface, so we can use staticcalls
• Again, rolling up the kill switch detection in AragonApp into canPerform(); to me this fits a lot better with the disabled / enabled model of an app
• Simplifying DAOFactory; I really liked how simple the old one was 😄

[facuspagnuolo]

Thanks for the review @sohkai! I already addressed/answered all your comments. Feel free to take another look if you want.

[facuspagnuolo]

The coverage task is failing because there are some tests that fail but only when measuring coverage. I've trying to debug it locally but it was completely cumbersome and I couldn't get to know what was going on. After that, I tried running solidity-coverage on using ganache-cli instead of testrpc-sc using frangio's fork and it passed. Definitely solidity-coverage is an outdated tool that has already been giving us several headaches 😕

[sohkai]

🎉 This is looking great; not sure how we should work around the coverage task for now (perhaps we can even just install frangio's fork 😄).

One thing I was looking into earlier (and unfortunately dropped) was doing some investigation into why the gas overhead was so high for checking the killswitch (27k without a SSTORE is quite high!). It'd be nice to see if there was anything obvious to make this a bit lower.