Return an accurate floating point sum of values in the iterable. Assumes IEEE-754 floating point arithmetic. This function has been overriden from math.fsum to work element-wise on iterables