Consider unifying the creation of covariance matrices for our test suite

[BjoernLudwigPTB]

Inconsistent ways of creating covariance matrices in the test suite
We have several different ways of creating those matrices for our Monte Carlo sampling and other purposes. Those are differently well thought through.

Unify the creation of covariance matrices in the test suite
We should put time and effort into finding a suitable way for creating such matrices and cleanly implement this. The main focus should lie on robustly creating suitable matrices for all the required cases.

Additional context
One disucssion dealing with this arose here