Add option to return NaN if window contains NaN for medfilt2d, medfilt

[nikkopante]

Is your feature request related to a problem? Please describe.
I am porting a script from Matlab to Python. Matlab's version of 2D Median Filter assigns an element to NaN when being computed and a window contains NaN.

Describe the solution you'd like
Possibly an additional argument, Boolean if NaN must be considered or ignored in the evaluation.

Describe alternatives you've considered
The silx library has this option for their [medfilt2d]. (http://www.silx.org/doc/silx/dev/modules/math/medianfilter.html#silx.math.medianfilter.medfilt2d). However the results are not accurate as SciPy's or Matlab's. I also tried ndimage.median_filter. However, it also ignores NaN and results we'rent as close as Matlab's.

Additional context (e.g. screenshots)
In Matlab or GNU Octave (Open Source)

pkg load image;
m = [1,2,3,4,5;1,2,3,4,5;1,2,3,4,5;1,2,3,4,5;1,2,3,4,NaN];
fltrd = medfilt2(m, [5 5])
fltrd =

     0     0     1     0     0
     0     1     2     2     0
     1     2     3     3     3
     0     1     2     2     0
     0     0     1     0   NaN

>>

In Python:

from scipy.signal import medfilt2d
import numpy as np
m = np.array([[1,2,3,4,5],[1,2,3,4,5],[1,2,3,4,5],[1,2,3,4,5],[1,2,3,4,np.nan]])
fltrd = medfilt2d(m, 5)
fltrd
array([[0., 0., 1., 0., 0.],
       [0., 1., 2., 2., 0.],
       [1., 2., 3., 3., 3.],
       [0., 1., 2., 2., 0.],
       [0., 0., 1., 0., 0.]])