Raise an error if gradient of QNode returning qml.sample() is computed

[eddddddy]

Feature details

Right now PennyLane allows computing the gradient of QNodes that returns qml.sample(). This is illustrated by the below code:

def print_sample_grad():
    dev = qml.device('default.qubit', wires=1, shots=10)

    @qml.qnode(dev, diff_method='parameter-shift')
    def circuit(x):
        qml.RX(x[0], wires=0)
        return qml.sample(wires=0)

    x = np.array([0.1])

    print('samples:', circuit(x))
    print('sample gradients:', qml.jacobian(circuit)(x))

>>> print_sample_grad()
samples: [0 0 0 0 0 0 0 0 0 0]
sample gradients: [[-0.5]
 [ 0.5]
 [ 0.5]
 [ 0. ]
 [ 0.5]
 [ 0.5]
 [-0.5]
 [ 0.5]
 [ 0. ]
 [ 0.5]]

This is unintuitive behaviour since the gradient of a sample doesn't really make any sense.

Instead of silently failing and producing nonsensical gradients, an error should be raised if the user tries to compute the gradient of a QNode that returns a sample. Ideally, the forward pass would still be supported.

Implementation

No response

How important would you say this feature is?

1: Not important. Would be nice to have.

Additional information

No response