Handle extra columns in taxonomy table better

[brendanf]

Ideally, the taxonomy table passed to phylotax() or lca_consensus() should be able to contain arbitrary extra columns, and those values should be propagated into the output as long as they have a single unique value for each label.