Allow skipping bad records instead of raising exceptions

[lbeltrame]

Due to the way bad records are handled currently it is really hard to identify what is the cause of an error (long SQLAlchemy backtraces aren't helpful) and makes it impossible to actually generate a database (normally I'd fall back to GEMINI, but I need hg38).

There should be an option of skipping bad records (off by default) when loading VCFs: not optimal, but at least would allow a database to be generated.

[brentp]

I think it would be better to improve the error message and/or provide a tool that will find these bad records.
If you have a single-line VCF that generates an sqlalchemy error, then I'll use it to work on improving the error messages.

[lbeltrame]

Examples are records without gnomad_af data, which cause a sqlalchemy error:

sqlalchemy.exc.StatementError: (sqlalchemy.exc.InvalidRequestError) A value is required for bind parameter 'gnomad_nfe_af', in parameter group 1

The parameter changes every time, because dictionaries are non ordered and the record is (excerpt):

 u'gnomad_afr_af': u'', u'gnomad_nfe_af': u''

[lbeltrame]

On second thought, what would be useful would be printing the actual VCF record, so people can at least locate the offending line easier.