Do not warn about unknown states if ancestral allele is "N"

[hyanwong]

In #960 @benjeffery said:

This has made me realise the vcfzarr spec should probably be more explicit that zero-length alleles are not allowed. It is currently there as implicit as it states "" is a fill value.

In that case, I presume that we can always assume that an ancestral allele of "" means "intentionally marked as unknown", so that we can recommend that people use that to mark sites of unknown ancestral allele. Furthermore, I suspect that means we can exclude such alleles from the "warning" output that normally says:

An ancestral allele was not found in the variant_allele array for the {tot} sites ({frac_bad * 100 :.2f}%) listed below.

?

[hyanwong]

Revised in the light of discussions in #963 so that "N" should be the "deliberately unknown" state. We still want to treat "" as unknown, but this is more likely to be accidental.

[benjeffery]

Yeah, I think you're right that we shouldn't warn on N