improve test error output

Currently, the tests check whether the expected output matches the actual output, but we don't give much feedback.

To improve the output, we could actually run a `git diff --exit-status expected.out actual.out` and report that