Measure correlation between posterior ensembles and optimal posterior ensemble, not just sample means and truth.

[Blunde1]

In Adaptive.ipynb it is stated

Correlations between true parameters and solution means
A more sophisticated way to measure fit might be to use Kullback–Leibler divergence.
Here we simply look at the correlations of the means.

We know what the optimal posterior update of the prior ensemble is. Let's compute this optimal (no sampling error) transport and measure correlations between full ensembles.