Make COUNT_ALIGNMENT_DUPLICATES use more clever parent selection.

[harmonbhasin]

Currently, our duplication analysis using Bowtie2 alignments assigns the parent of the duplicate reads as the one with the highest average phred quality score. However, as Mike pointed out there are better solutions. In the future, we may decide to implement this.

[willbradshaw]

I expect this will shortly be rendered moot by #122, which will involve removing deduplication from the core pipeline. However, a very similar issue might be inherited by whatever downstream script or pipeline takes on that functionality.

Moving this down to priority 3 for now.