Create VcfReader class or standalone function which takes INFO key-value pairs and reconstructs VRS Alleles

[theferrit32]

We let people annotate VCFs with info we claim is enough to reconstruct the VRS Alleles for each record, but we don't provide the code to do that.

Providing this would be useful to users, would prove that the annotations do work, and would be useful in testing.

If we go the VcfReader route we should provide a generator interface over the records which returns the original pysam records, along with the list of VRS Allele objects constructed from the annotations.

If the input is not already annotated, it should fail.

[theferrit32]

A thing to keep in mind (raised by @jsstevenson) is that pysam serializes None and "" to a . char in the VCF file. And deserializes . to None. There are cases like .sequence where None and "" do mean different things. So to reverse . we do actually potentially need other context to know if we should interpret it as None or "".

For LiteralSequenceExpression, the VCFAnnotator will never result in a .sequence="", because deletions become ReferenceLengthExpressions now. But you can manually create a non-normalized LiteralSequenceExpression with .sequence="", and annotate a VCF with that value. I'm not sure we need to care about this because this would be the user doing non-standard and explicitly not-recommended behavior.