Linting (in particular the tests)

[florian-huber]

When switching on the linter in #273 I came across quiet some linter issues. Some are solved in #273, and some I assume we don't want to "fix" (e.g. having many arguments is part of the deep learning game...).
But many of the tests also caused linter issues and in some cases I was wondering if we weren't missing some additional asserts etc.
So, maybe we can:

• set the linter profile a bit stricter and address the raised warnings
• Check the actual test coverage, e.g. by adding SonarCloud?
• Check the unit tests and see if they contain the right asserts