update sign convention of Pauli Y

[jrm874]

This (very minor) PR attempts to fix the discrepancy in the definition of the Pauli Y operator.

The standard definition is Y = [[0, -1j], [1j, 0]].

In the current version, when projecting the Y operator into a subspace spanned by [[False], [True]], one obtains the matrix [[0, 1j], [-1j, 0]] = -Y.

I think that the tests did not cover it because of two reasons:

1. In the tutorials we test the Heisenberg model, which contains products of Y operators acting on pairs of qubits -> the sign cancels-out.
2. In the unit tests we only check for the smallest eigenvalue of a single Pauli operator projected into the subspace. This eigenvalue is always -1 regardless of the definition of the sign of Y.