Handle unmarked duplicates where mates have MQ = 0

[davidbenjamin]

PCR duplicates may not be marked if their mates all have mapping quality zero and are assigned to different places. This causes false positives in Mutect2.

[jamesemery]

@davidbenjamin If you aren't aware there is a proposal to handle this case better in MarkDuplicates floating around in Picard. If you haven't already expressed your opinions to the relevant people it may be worth leaving a comment on the discussion broadinstitute/picard#1285 (comment).

[davidbenjamin]

Thanks @jamesemery!

[davidbenjamin]

Closed by #5734.