Update with alterantive read.bam.tags function

[fferrari]

Hi @pkharchenko, we came across an issue with BAM files containing paired end reads.
The current read.bam.tags() function has an "if" conditional statement checking for paired end reads in the input BAM
if(any(bitwAnd(bam$flag,0x1))) { # paired-end data

However, there are 2 problems with this

1. the code inside the if statement is taking only reads mapped on the positive strand
   posvi <- which(strm==1); rl <- list(tags=tapply(posvi,bam$rname[posvi],function(ii) as.numeric(bam$pos[ii])), flen=tapply(posvi,bam$rname[posvi],function(ii) as.numeric(abs(bam$isize[ii]))))
   and this will not be compliant with cross-correlation and peak calling algorithms which expect instead reads distributed on both strands

2. the code handling quality scores is outside the if statement
   rl <- c(rl,list(quality=tapply(1:length(bam$pos),bam$rname,function(ii) bam$mapq[ii])))
   thus resulting in an output taglist object with twice as many quality scores as compared to actual reads positions

I'm proposing an alternative function in a pull request.
The alternative function will randomly pick only one of the reads in the pair when both are aligned, thus resulting in a taglist object with characteristics similar to a single end BAM file. however, I am not sure this was the intended purpose of the original code handling BAM files with paired end reads.

[MateuszChilinski]

Hey,

in version 1.16, when parsing pair-read data, we encountered an error where tag.scc was returning -inf and inf from some max/min functions because all tags were positive. This pull request fixes that, so I'd suggest merging to master. I think the fix presented is also appropriate. Good job! :)

For compatibility, I'd rename f_read.bam.tags to read.bam.tags though.