Fitting null with perfect separation

[johnlees]

This dataset fails: http://klif.vet.uu.nl/pyseer/genomes/

[mgalardini]

Hi, yes, it is true that the only way to make this work is to flip any random phenotype. Maybe we can add a warning message when something like this happens?

A side note: the phenotypes.tsv file is missing the header. The null fit still fails with it though.