Fix batch_execute output type for QNodes with multiple outputs

[antalszava]

Context

One of the demonstrations fails due to a change of output type when using the qiskit.aer device.

With the latest batch_execute addition, this device may output results in a different format than other devices (e.g., default.qubit).

from pennylane import numpy as np
import pennylane as qml

dev2 = qml.device("qiskit.aer", wires=4)

@qml.qnode(dev2, interface="torch")
def qnode(params):
    qml.RX(params, wires=0)
    return qml.expval(qml.PauliZ(0)), qml.expval(qml.PauliZ(1)), qml.expval(qml.PauliZ(2))


qnode(0.3)

[tensor(0.9512, dtype=torch.float64),
 tensor(1., dtype=torch.float64),
 tensor(1., dtype=torch.float64)]

whereas it should output

tensor([0.9746, 1.0000, 1.0000], dtype=torch.float64)

The reason why this causes is an error in the demo is that we're using torch.stack in the QNodeCollection class on the results of the underlying QNodes. However, with the latest output type, we may end up having a list in the output, on which torch.stack will fail.

Internally, the issue is a missing conversion to np.array, that has been done previously when using the default execute pipeline.

This can be showcased by comparing the outputs of executing the same tape using the default.qubit and the qiskit.aer devices:

from pennylane import numpy as np
import pennylane as qml

with qml.tape.QuantumTape() as tape:
    qml.PauliX(wires=0)
    qml.expval(qml.PauliZ(wires=0)), qml.expval(qml.PauliZ(wires=1))

dev1 = qml.device("qiskit.aer", wires=4)
print(dev1.batch_execute([tape]))

dev2 = qml.device("default.qubit", wires=4)

print(dev2.batch_execute([tape]))

[[-1.0, 1.0]]
[array([-1.,  1.])]

Changes

Converts the output of QubitDevice.statistics to a numpy array to resolve the issue.

[codecov]

Codecov Report

Merging #163 (5424047) into master (63f2efa) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master      #163   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            7         7           
  Lines          313       314    +1     
=========================================
+ Hits           313       314    +1     

Impacted Files	Coverage Δ
pennylane_qiskit/qiskit_device.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 63f2efa...5424047. Read the comment docs.

[rmoyard]

Thanks @antalszava I tried different cases and it is working! Just left one comment 💯

[antalszava]

[sc-11994]