qQuantileNormalizer with unweighted source and weighted destination

[Peter]

The qQuantileNormalizer::operator has four cases:

1. unweighted source and unweighted destination
2. unweighted source and weighted destination
3. weighted source and unweighted destination
4. weighted source and weighted destination

Currently if either of the ranges is weighted, i.e., case 2-4, the weighted version of normalize is called. The question is: is this really needed for case 2?

Only time destination weights are used is when they are copied from source via function copy_weight_if_weighted. If we include a call to that function also in the unweighted version of normalize (which is faster), we could let that function handle case 2 as well. When the destination is unweighted, copy_weight_if_weighted is a no-op, so the change will not effect the speed of case 1.

[Peter]

(In [3572]) fixes #874