Fix filtering of paired-end reads

[csoneson]

(Moved here from slack - thread from April 5, 2023 contains a link to a reproducible example)

• The problem first manifested when running qAlign with an auxiliary fasta file for a paired-end sample, using Rhisat2

Error in checkForRemoteErrors(val) : 
 one node produced an error: Error processing sample XX_R1_001.fastq.gz : The order of unmapped paired-end sequences in bamfile is inconsistent at 57577483-th alignment.

• The error message seems to come from

  QuasR/src/extract_unmapped_reads.c

  Line 102 in cd52416

  Rf_error("The order of unmapped paired-end sequences in bamfile is inconsistent at %i-th alignment.\n", count);

  - the bam file above contains a lonely mate at the position indicated in the error message.

• The "unmapped" mate is probably multimapping (and thus removed by QuasR), while the "mapped" mate maps uniquely (and thus is still reported).

• Solution: For overmapped reads in paired-end experiments, the logic should remove them in pairs rather than as single reads.