Formula-error in MACC

Hi,

the formula of MACC is wrong in s2dverification because its summing the square-root of variances. The correct formula should average the variances first before computing the square root (http://journals.ametsoc.org/doi/pdf/10.1175/1520-0442(1992)005%3C1346:TSOERE%3E2.0.CO%3B2). I will make a new branch to correct this and try to give an illustration how big the impact is.

@nmanubens manubens @cprodhomme @lpcaron @fmassonnet @vsicardi @eexarchou @mmenegoz @nfuckar @vguemas