Merit class UncertaintyCorrelation yields nan for constant predictive uncertainties

[mrupp-citrine]

When calling UncertaintyCorrelation with predictive distributions that have constant uncertainty, value varSigma (line 131 in Merit.scala) is zero, leading to denominator being zero. As the numerator is also zero, not-a-number (nan) is returned.

This behaviour seems undesirable. I suggest throwing an exception to signal the problem instead of silently inserting nan into calculations.