Add more lenient "bgn"-style imputation output

[rkwalters]

Filtering imputation output on quality of best-guess genotypes tends to be stricter than providing dosages. Providing a more lenient set of output, analogous ricopili's "bgn" best-guess, alongside the current filtered results may be useful. Alternative is to output dosages, but that lacks analysis options at the moment.

Especially valuable for SNPs that may be on the edge of imputation quality that pass dosage QC and produce interesting GWAS results for case/control data, but "disappear" in family data since they don't pass best-guess QC. (Note: this is a documented occurrence, not a hypothetical)