Add multithread support for some CPU processing classes

Some processing classes like CCDCorrection and UnsharpMask are quite slow.
However they are easy to parallelize with native python multi-threading, as the underlying processing function release the GIL.

from nabu.preproc.ccd import CCDCorrection
from multiprocessing.pool import ThreadPool

ccd_cor = CCDCorrection(img_shape)
threads_pool = ThreadPool(6)

def img_median_filter(image):
    out = ccd_cor.median_filter(image)
    # In-place, can easily be extended to out-of-place
    image[:] = out[:]
    # map() creates a list containing the result from each thread
    # so returning the processed image ends up duplicating all the images stack memory.
    # Therefore, it is very important to return None    
    return None

threads_pool.map(img_median_filter, img_stack)