fix for #252

[scopatz]

This is a fix for this negative step size issue, since hsize_t on many platforms is now an unsigned long.

[avalentino]

@scopatz also applying the patch it seems that revere iteration still doesn't work (see 255e669).
The fix seems to be not immediate. Also we should make a decision about gh-44 that seems to be related to table iteration and implies an API change).

[scopatz]

So this fix that I put in works for when step is -1 but not when it is < -1. Thus the test case I was working with (below) passes, but the one that you put in the test suite doesn't since step = -5 there.

import tables

class IndexRecord(tables.IsDescription):
    frame = tables.Int32Col(pos=1)
    key = tables.StringCol(itemsize=40, pos=2)     

h5 = tables.openFile("trace.h5", 'w')
index = h5.createTable(h5.root, "index", IndexRecord)


index.append([(0, 'A')])
index.append([(1, 'B')])
index.append([(3, 'C')])
index.append([(4, 'D')])
index.flush()

index.cols._f_col('frame').createCSIndex()
for row in index.itersorted('frame', step=-1):   #<-crashes here!
    print row['frame'], row['key']

from http://stackoverflow.com/questions/16407274/reverse-iterate-pytable-with-itersorted-and-negative-step-produces-overflowerror

[avalentino]

Merged with commit 7c2093a.