==============
Highlights
----------
- When training ChromBPNet bias models from the command line and no loci
are provided for training the bias model, the default is changed to inherit
from `negatives` instead of `loci`. Colloquially, when the user provides a
set of negatives and peak loci to train ChromBPNet, the default is now that
the bias model will be trained on the negatives instead of incorrectly on
the peaks.
- When calculating the maximum number of reads a negative region can have
when training the bias model, changed from using the minimum number of reads
in the peaks to the 1st quantile of reads because it is more robust.