Robust estimate of sample/cosmic variance

[steven-murray]

We've mentioned a few times that we need to be able to robustly estimate the uncertainty we expect from running multiple EoR/FG simulations. This should play into how we "accept" a test, setting the inherent level of uncertainty.

This issue is a place to discuss the details of this question, and also to propose ideas of future explicit tests.

We have already decided that such tests are best suited to post-H1C.