supports excluding reads by a tag's value, with tests

[bwlang]

No description provided.

[brentp]

hi, the implementation looks fine, though I'd rather move the spec.find(':') to before the alignment processing so it's only done once.
I am hesitant to add since it doesn't seem to be widely needed. What's the use-case for this?

[bwlang]

hi... thanks for getting back!
my application is for measuring coverage of "optical duplicates" these are encoded only on a tag (the flag mixes both PCR and optical duplicates). I wanted to see if they cluster up at particular loci. Once could do this with samtools filtering ahead of mosdepth at the cost of an extra bam copy (in this case 100+G). There may be other tag only encodings in e.g. methylation sequencing bams written by bismark etc. I'll make the change you suggested and let you decide whether this is worth a few more lines.