CSI Implementation for Larger Genomes

Description of feature

Hi,

I've been helping my coworker run this pipeline, but we hit an error with SAMTOOLS_INDEX, which I see isn't that uncommon. Using ext.args, we were able to resolve the error, but only a partial run followed. The last step that ran at that point was MERGED_LIBRARY_MARKDUPLICATES_PICARD, so I was able to git clone and modify the atacseq.nf at line 480 to read
.join(MERGED_LIBRARY_MARKDUPLICATES_PICARD.out.csi, by: [0])

This change allowed for more of the pipeline to run, but it still did not complete. This time, the pipeline was halted at the bam_filter_bamtools.nf subworkflow. I modified line 89 to read
bai = BAM_SORT_STATS_SAMTOOLS.out.csi.mix(SAMTOOLS_INDEX.out.csi) // channel: [ val(meta), [ bai ] ]

These changes allowed for the pipeline to complete using the csi index, and I think it'd be useful to implement these as more global changes. If I didn't have a thorough Nextflow background, I don't think we would be able to use the complete pipeline for this species, so I think making adjustments to allow for easier csi usage would benefit the community as a whole! Let me know if you'd like any additional info!

Start with the CSI-related channel joins in atacseq.nf around line 480 and the bam_filter_bamtools.nf subworkflow around line 89. Trace how BAM indexes are passed through these steps and identify the other pipeline points that assume BAI indexes. Done means the complete pipeline runs for larger genomes using CSI indexes without manual file edits.