ResetChannel does not do in-place modification of density matrices with qibojit

[jf-kong]

When using qibojit as backend, the initial state is typically modified in-place. This generally works for simple gates and with density matrices. However, ResetChannel does not seem to modify the initial density matrix in-place, as the following example shows. Is this the intended behaviour?

import numpy as np

import qibo
from qibo import gates
from qibo.models import Circuit

qibo.set_backend("qibojit", platform="numba")

circuit1 = Circuit(2, density_matrix=True)
circuit2 = Circuit(2, density_matrix=True)

# Create a simple circuit
circuit1.add(gates.H(0))
circuit1.add(gates.CNOT(0, 1))

# Create a second circuit that differs from the first only with a ResetChannel
circuit2.add(gates.H(0))
circuit2.add(gates.CNOT(0, 1))
circuit2.add(gates.ResetChannel(1, [1.0, 0.0]))  # The only difference

# Create two 4 x 4 identical density matrices
initial_state1 = np.array(
    [[0.5, 1, 0, 1], [1, 0.5, 0, 0], [0, 0, 0, 0], [1, 0, 0, 0]], dtype=np.complex128
)
initial_state2 = np.copy(initial_state1)

# Now execute the two circuits
final_state1 = circuit1(initial_state1)
final_state2 = circuit2(initial_state2)

# The intended behavior is that the initial states are modified in place (?)
# As such, the initial states should be close to the corresponding final states
# Test for closeness of the corresponding states
assert np.allclose(final_state1.state(), initial_state1, atol=1e-6)  # This passes
assert np.allclose(initial_state2, initial_state1, atol=1e-6)  # This passes when it shouldn't
assert np.allclose(final_state2.state(), initial_state2, atol=1e-6)  # This fails