speed up weighted qQuantileNormalizer::normalizer

[Peter]

related to tickets #535 and #536

There is a code block in weighted version of normalize that was added to fix bug #535. With the new versions of sort_index (available since yat 0.13) this is no longer needed. Instead, the line

utility::sort_index(utility::data_iterator(first),
                    utility::data_iterator(last), 
                    sorted_index);

could be replaced with something like

utility::sort_index(utility::data_iterator(first),
                    utility::data_iterator(last), 
                    sorted_index, less_nan());

This will avoid copying the data before running sort_index. It will change behaviour in some odd cases (like input with NaN and finite weight), but not in any sensible cases, AFAICS.

[Peter]

(In [3571]) closes #875

With new version of sort_index (since yat 0.13) allows sorting ranges of DataWeight?, which behaves as desired, use that function rather than copying input range.