The side effect of this function is to drop the rows (individuals)
with unclearly identified genotypes, i.e., with at least one allele
among na.alleles.
Other variables in the data table are eventually checked and levels
with no observation (e.g., population) are dropped.