support extracting skip-mers rather than kmers from reads

[notestaff]

As described at https://www.biorxiv.org/content/10.1101/179960v2 , extracting "skip-mers" (kmers that skip every 3rd base) can improve matching in coding regions. An option to extract such skip-mers rather than normal k-mers from input reads would be useful.

Related: #142