standardize canonical coordinate conventions in quilc

[macrologist]

The coordinates used in PCS, which I think are what's used here, have a goofy normalization condition that causes its canonical gates to span a triangular prism rather than a tetrahedron. ... inject an extra step here to re-scissor its output into the tetrahedron you expect ... I don't think much (anything?) in quilc depends on that alternative convention, and it was definitely a flub on our part not to do that from the start.

Originally posted by @ecpeterson in #868 (comment)