Provide multiprocessing option for large classification jobs

[GoogleCodeExporter]

Large datasets (e.g., agemap H & E stain ~40,000 samples) take an extremely 
long time in Pychrm to do train/test split and classify operations. Euclidean 
distances can be calculated in a parallellized way, e.g., one processor can to 
all the samples from a given class.

This would entail exposing the samples in the FeatureSet.data_matrix to C++. A 
C++ implemented, Python-wrapped wndchrm classify option would also speed up 
computation.

Original issue reported on code.google.com by [email protected] on 18 Jan 2013 at 9:33