Calculate the arithmetic mean of a double-precision floating-point strided array, ignoring NaN values and using pairwise summation.