Prevents negative variance (mathematically impossible) in Variance
indicator.
Since variance is used to calculate standard deviation, we found that
negative variance yield double.NaN and, consequently, an arithmetic
overflow when we convert NaN into decimal.
Closes#2672