negctrl expansion emits the same QuantumGate object twice, aliasing operand nodes #350

Description

The negctrl modifier expansion in visitor.py places the same QuantumGate object at two positions in the emitted statement list, rather than two equal copies:

negs = [
    qasm3_ast.QuantumGate([], qasm3_ast.Identifier("x"), [], [ctrl]) for ctrl in negctrls
]
result = negs + result + negs

Every transformation that rewrites qubit indices in place (remove_idle_qubits(), reverse_qubit_order()) therefore reaches the same operand node more than once.

Reproduction

from pyqasm import loads

qasm = """OPENQASM 3.0;
include "stdgates.inc";
qubit[3] q;
negctrl @ x q[0], q[1];
"""
m = loads(qasm)
m.unroll()
print([id(s) for s in m._unrolled_ast.statements if hasattr(s, "qubits")])
# [4362757584, 4362758224, 4362757584]  <- statements 0 and 2 are the same object
m.reverse_qubit_order()  # KeyError: -2

remove_idle_qubits() on the same circuit survives only because of the visited_node_ids guard in Modules._remap_qubits; removing that guard raises KeyError: 0. Nothing in the test suite covers either path.

Scope

Pre-existing, not a regression. #335 fixed the equivalent aliasing for decomposition-emitted statements (maps/gates.py constructors and Decomposer), but deliberately did not touch the modifier expansion path.

Suggested fix

Build a second negs list for the trailing position rather than reusing the first, and run the control operand through fresh_qubits() so the two x gates own distinct operand nodes. Add a regression test asserting no shared statement or operand nodes after unroll() for negctrl @ x, and that reverse_qubit_order() succeeds on it.

Found in review of #335 (M2).