clarify that genetic_relatedness includes self comparisons

[petrelharp]

As discussed in #2888, usually "genetic relatedness" includes the self comparisons. (This makes sense if (a) you're computing the covariance matrix of additive traits, (b) if you want to know the future inbreeding of a pop started from those individuals. However, it makes less sense if it's an estimator of a quantity of a larger group of individuals, since then the estimator depends on the sample size. So, this should probably be optional.)

[petrelharp]

Ah - actually, it appears that we do already include self comparisons. Perhaps we should clarify this in the docs, since most other stats do not. (We could make this an optional thing, but this is less essential.)