Bug in TensoredConfusionMatrices dosctrings

[alejogoogle]

Describe the issue

First is said that the CM columns correspond to the same preparation (i.e. column 0 corresponds to having prepared 0).

and just after, in the 2Q case, now the rows correspond to the same preparation (row 0 corresponds to having prepared 0)

I think the 2Q case is right, but the 1Q one is wrong? It is a little confusing or am I missing something?

[mhucka]

@alejogoogle Thank you for reporting this and opening a ticket. Yes, you're correct: the 1-qubit case is wrong here. Cirq's implementation elsewhere consistently uses the row = prepared, column = observed convention, but the 1Q here is flipped.