[bug] fix the confusing order of elements in SNPSample.shape, fixes #76
Previously the SNPSample.shape attribute was a tuple ordered (n_snp, n_indiv)
. I don't know why I did it that way in the first place. I
think originally my thinking was "n_snp" was somehow "more important"
and should come first. But really there's no rational reason for this,
and it's only deeply confusing since it the opposite sense of the shape
of the SNP matrix tensor itself.
In general SNPSample.n_snp and SNPSample.n_indiv should be preferred in most code, as this is more explicit.
Recent versions of PyTorch now also have prototype support for named dimensions on tensors: https://pytorch.org/tutorials/intermediate/named_tensor_tutorial.html