Implement mask for bilinear interpolation

[kif]

Provide a mask in the numerical kernel in charge of bilinear interpolation.
Returns Nan when no underlying pixel is unmasked. Interpolates approximately then 1, 2 or 3 underlying pixels are masked.

This is part of #3285

[kif]

In [1]: %pylab
Using matplotlib backend: TkAgg
Populating the interactive namespace from numpy and matplotlib

In [2]: import fabio, silx.image.bilinear

In [3]: f=fabio.open("/mnt/data/eiger/Eiger_raw/x1_insu_0p05d_0p05s_0p01beam_data_000002.h5")

In [4]: c = silx.image.bilinear.BilinearImage(f.data, mask=f.data==65535)

In [5]: plot(c.profile_line((0,0), (1000,1000)))
Out[5]: [<matplotlib.lines.Line2D at 0x7fc564c01630>]

[kif]

[vallsv]

This makes me think what about an option to return nan instead of 0 when there is a mask.
The rendering of the profile could be more understandable.

[kif]

This sounds like a good idea. I did not even notice.
if cnt is null, set the value to NAN.