handle Nan values in signtest.m

[AvanishSalunke]

1. when NaN values were passed, the function never filtered them out :

XY_diff = x(:) - my(:);
NO_diff = (XY_diff == 0);
XY_diff(NO_diff | isnan (NO_diff)) = [];

so we got differences in the output
matlab :

>> x = [1, 2, 3, 4, NaN, NaN, NaN];

>> [p, h, stats] = signtest(x)


p =

    0.1250


h =

  logical

   0


stats = 

  struct with fields:

    zval: NaN
    sign: 4

octave :

octave:5> [p, h, stats] = signtest(x)
p = 1.0000
h = 0
stats =

  scalar structure containing the fields:

    zval = NaN
    sign = 4
    

2. The null hypothesis is rejected also when p value is equal to the alpha value.