Improve CSV-based comparison?

[CSchoel]

We could take a look at the algorithm used in modelica-tools/csv-compare to get ideas how to make the comparison of simulation outputs a little less strict. Then again, the question is if we want to be less strict. I think it is at least an option to consider, but we would maybe need some additional checks for event localization.