Detect when i5 indices need to be rev-comp'ed

[tomkinsc]

A task, revcomp_i5, has been present for a while to reverse complement i5 (barcode_2) indices in sample sheet files. The illumina_demux functionality should detect when this is likely needed, and when it is the i5 indices in the supplied sample sheet(s) should be automatically rev-comp'ed before demux is run again with the modified sample sheets.

This should be possible by using the IlluminaBasecallsToSam debug options FIRST_TILE=1 and TILE_LIMIT=1 to restrict demultiplexing to a single tile, and then inspecting the count of reads in the demux metrics file written by picard: rev-comp i5 seqs if [(sum of counts in READS for samples) / (count of reads in "other" row, NNNNNNNNNN-NNNNNNNNNN)] > some low threshold.

This should have a guardrail to prevent an incorrect sample sheet from being rev-comp'ed and re-demuxed ad infinitum if it is incorrect for some other reason.