Update SICEv1.0

[BaptisteVandecrux]

On the optimization, the major updates are:

• Calculation of the solar fluxes outside of the loop (same for all pixels)
• Correction of ozone/water scattering and O3_SICE calculation is now done as matrix operation
• Calculation of all geometrical parameters, atmospheric and clean snow characteristics (among which D and r0) done as matrix operation
• Calculation of the polluted snow characteristics done on per-pixel basis (needs iterative solver) but optimized using numpy.vectorize
• Calculation of planar and spherical albedos done as matrix operation
• Bands for which the retrieval was failing were previously linearly interpolated using the neighboring bands, these cases are now left as NaN
• Integration of spectral albedo over three intervals (among which SW for BBA) done on pixel basis but optimized using numpy.vectorize

The new script is now between 1.25 and 5 times faster.

Note that the slow part of the script is when solving for dark pixels. Applying a land/ocean mask at the scene level would speed up even more the script. At the moment land/ocean masking is done in dm.sh after sice.py was run. Once the cloud mask comes from idepix, we can run sice 1x/day rather than 1x/scene (see #19).

On the equations that are being used, the major update is that the BBA calculation for polluted snow now uses the BOA reflectance instead of the TOA reflectance as in the previous version. Alex updated that in the fortran code in December. Expect different performance at the PROMICE stations.

[mankoff]

I'm not sure why there are such large changes involved in this, but I'm going to merge it.