Lenient mode tweaks

[remiolsen]

The decision to flip indices needs to be weighted by the amount of reads with "unknown" barcodes. For instance currently it can decide to flip I7 if it will go from 1 read to 10 reads demuxed, whereas there might be 100k reads with unknown.

Edit: for my own reference, see run 20240118_1039_MN19414_ASG790_ae66b218

[remiolsen]

This is a bug! Output from lenient mode for a pol (internal note: run on flowcell 20240125_1233_MN19414_API303_1947636a):

DEBUG:anglerfish: Matched 0 reads, unmatched 188763 reads
DEBUG:anglerfish: Matched 0 reads, unmatched 181962 reads
DEBUG:anglerfish: Matched 198071 reads, unmatched 12537 reads
DEBUG:anglerfish: Matched 2 reads, unmatched 191222 reads <--- Lenient mode selected this one!!!

This shows out of the four possible orientations: original, revcomp i7, revcomp i5 and revcomp both it did not make the obvious choice which would yield many more matched reads.