This required, for some of the competitions, to run the data through
dos2unix, which changed the shas.  But also, at the etl library level,
we needed to remove the unicode <feff> character that sometimes gets
added to the start of the csvs.