CI test for noise realisation
Related, @gregory.ashton and I were talking, and we both think we should have a CI test that confirms the noise realisation returns something sensible. Greg's idea was for the calculation of a single noise realisation, and then confirming that 50% (or so) of points are within one-sigma of the ASD. (The reason for choosing 50% is just so that it breaks very rarely). Of course, we could do something more robust than this (e.g., calculating 1000 realisations and making sure the mean is close to the supplied ASD), but we also don't want it to be computationally burdensome, etc. Thoughts?
Moved this here because it did not relate to the other issue.