Consider adding a low-complexity sequence masking to genomes before building #96
Description
Description of feature
This can in some cases improve classification as it doesn't try to find the best place for unspecific reads (assuming a tool accepts such masked genomes).
It may also make databases smaller as it will remove reundant unspecific regions (e.g. by reducing the number of kmers)
E.g. with dustmasker
Idea from: https://github.com/khyox/recentrifuge/wiki/Centrifuge-nt#step-by-step-instructions