Wrong signs in PauliPolynomial propagation (maybe)?

[Ectras]

I think the signs in the PropagateClifford implementation of PauliPolynomial might be wrong.

Example: Propagating S through an X rotation

For instance, when propagating an S from left to right through an X rotation, we should pick up a minus sign (source):

However, in the code it's

fn s(&mut self, target: IndexType) -> &mut Self {
    let chains_target = self.chains.get_mut(target).unwrap();
    // Update angles
    let y_vec = chains_target.y_bitmask();
    for (angle, flip) in zip(self.angles.write().unwrap().iter_mut(), y_vec.iter()) {
        if *flip {
            *angle *= -1.0;
        }
    }
    chains_target.s();
    self
}

where s() on a chain is defined as

pub(crate) fn s(&self) {
    *self.z.write().unwrap() ^= self.x.read().unwrap().as_bitslice();
}

So for the example of a chain that is just a single X, no sign flip happens (because there is no Y) and only then the X is changed to a Y.

To make it more concrete, the following test code fails contrary to my expectations:

#[test]
fn pass_s_through_x() {
    let mut pp = PauliPolynomial::from_hamiltonian(vec![("X", 1.0)]);
    pp.s(0);
    let pp_ref = PauliPolynomial::from_hamiltonian(vec![("Y", -1.0)]);
    assert_eq!(pp, pp_ref);
}

Other cases

The sign behavior is reversed, for both s() and v().