Properties of qip.random.choi_matrix are not clear

[marcusps]

It is not clear what quantum operations are randomly drawn in the generation of random Choi matrices. They do not appear to be trace preserving, since tracing out the second subsystem does not lead to the identity matrix (although it does lead to a trace 1 matrix). Parhaps it leads to scaled projections are all, and compositions of that with other valid CP maps.