Add heterogeneity measures to Estimators

[tsalo]

It looks like DerSimonianLaird calculates Cochran's Q already, so we should be able to (1) extend it to calculate H2 and I2 and (2) return all three (or even just Q and I2) as outputs.

This is related to neurostuff/NiMARE#591.

[tsalo]

It looks like we actually have those already. They're just calculated in the MetaRegressionResults rather than the Estimator, which makes sense.