option to allow number=A fields with too few values

[snarfed]

we currently have a couple number=A INFO fields in our VCFs that often only have one value. right now, VT's validation is strict and doesn't allow these. it's right, of course, and we should either fix them or remove them from our VCFs...

...but in the meantime, i've been using a small hack to allow these, and put the value in just the first allele (again, incorrect, i know :P): snarfed@1aa4f4f . it's from before parts of bigquery_vcf_schema.py were refactored into processed_variant.py (cc @bashir2).

would you all consider maybe allowing this, or doing something similar, with a flag? maybe with --allow_malformed_records? i'm open to alternatives.

[snarfed]

(#79 and #101 are related)

[bashir2]

Have you considered using the --representative_header_file option. It seems to me that you can fix this problem by providing a VCF header that has Number=1 for those INFO fields. With this solution the field will be correctly added to non-allele INFO fields as well.

[snarfed]

good point! that could definitely work, at least for now. we actually already use it, i just didn't think to change its Number= for these fields. eventually we'd like to stop using a representative VCF file, though, so we can add new fields as in #83.