neoag: primary datasets take precedence over CEDAR/Neopep; trim blank cols from iedb_labeled

DESCRIPTION: label iedb_labeled as the ipred training set (#7)

Rework #9 into workhorse neoag DB (tested + candidates) + HLA population freqs

Add v1 data compendium: immunogenicity labels/benchmarks + similarity references

Add pMHC dataset: epitope-MHC binding from IEDB

initial commit