add feature "markdup support multiple input bams and argument file"

[lituan]

Hi,

smabamba is really fast doing markdup, yet it cannot support multiple input bams, which picard can.

we usually split multiple fq into small chunks, then align, then merge, then markdup.

with picard, we can do split, align, markdup, excluding merge step

so, if sambamba can support multiple input bams in markdup, it will help a lot

[pjotrp]

Good idea.

[bwlang]

would also be really nice to include an option to pass a file of filenames when paths are long or there are many bams.
Possibly related... adding a --sample #number_of_reads option to markdup would allow skipping another file (it would need to apply the sampling logic before dup logic to avoid reads marked dup in the output but the original read being filtered out)