Update CreateGermlines to be consistent with clean_imgtdb.py and use the first allele

There is a inconsistency between CreateGermlines and clean_imgtdb in how they select duplicated alleles. We have noticed this in mouse light chain genes (example IGLV2*01). clean_imgtdb uses the first occurrence of the duplicated allele to build the igblast database, but CreateGermlines uses the last occurrence to reconstruct the germline.