Manager refactor, part 1

[DingDongSoLong4]

The time has come for more refactoring.

This PR mostly deals with the initialization process. In general, the whole process is now a lot simpler and more straightforward.

The main function in cmd/stash/main.go is now a lot cleaner. A channel is now used to exit with an exit code, rather than just calling os.Exit in Manager.Shutdown. This means that everything can be properly shut down using defers.

I've also moved a few things around - the manager initialization code is now in a separate init.go file for example. And some renaming: config.Instance to config.Config and Manager.FFMPEG to Manager.FFMpeg.

And then a small fix to the Makefile after #4153, to prevent "nothing to be done" warnings even when specifying multiple flags-* targets (e.g. make flags-release flags-pie stash).

[WithoutPants]

Setup is broken with these changes. I'm getting a crash due to concurrent map read/write access in the viper code. If I then restart stash and retry the final setup step, it returns a fetch error. This may be an existing latent bug when trying to re-setup the system after it was created (since the config.yml was created prior to the crash).

[DingDongSoLong4]

I wasn't able to reproduce the crash myself, but I was able to find a data race which I assume was the cause with the help of -race. I was also getting the same data race warning on develop - the bug appears to have been there for a while.

And then the setup error I believe was simply that the error state in the Setup component wasn't being reset after clicking "Back" - so if an error occurs the first time it sticks around and you can't continue, even if no more errors occur.