Add non-paired reads filtering stage to read classifiers that can process paired reads

Some reads classifiers such as Kraken can process paired reads. But the input reads can have no pair because of preprocessing, e.g. read can be filtered out by quality. Kraken has an option to check reads names during work, but in case of unpaired read it stops work.
Add an additional preparation task to remove reads without pair from the input files. Add a parameter to enable/disable reads filtering to the classification elements. The default value should be "True". The parameter should be visible only for paired reads processing mode.