Mass weighting, and other weightings, of RMSE and CORR

[painter1]

Some time today, the master branch will see @mcenerney1 's computation of the RMS error RMSE and correlation CORR between model & obs (or model & model, etc.). These are computed with genutil functions. Both RMSE and CORR involve averages, and thus should be weighted as the mean calculations are weighted - sometimes mass weighting, sometimes area weighting. I do not know what weightings the genutil functions use, but we can be sure that is never is mass weighting.