Mapped counts in review decisions are wrong

[donkirkby]

When a sample is remapped, the QAI upload reports the maximum number of reads mapped to each seed. However, that doesn't make sense when a seed maps fewer reads in later iterations. It should just report the final number of reads mapped to each seed.

As an example, see sample 88200A-HCV_S61 from the 16-Jun-2017.M04401 run.

A similar bug is that V3LOOP samples report the wrong number of mapped and raw reads in the QAI upload. They should be adding in numbers from the g2p files. An example of that is sample 69374A-V3-2-V3LOOP_S2 from the 30 Jun 2017.M04401 run.

• fewer reads in later iterations
• V3LOOP read counts