Why not append None to all reported SampleData allele lists?

[hyanwong]

In various places in tsinfer we try to work out if a site contains missing data, and if so, we append a None to the alleles list, so that if we access into the list, we get a None out.

Sorting out the logic of whether a site has missing data leads to some extra complex code. Why don;t we simply add a None to all allele lists, so that the number of alleles is always the size of the list minus 1? What are the downsides to that?