[BUG] Gradient verification fails if there is a `Prod` operation somewhere in the circuit

[josh146]

For the following circuit, gradient verification fails:

@qml.qjit
def f(params, x):

    @qml.qnode(dev, diff_method="parameter-shift")
    def circuit(params, x):
        qml.IQPEmbedding(x, wires=range(3), n_repeats=1)
        qml.StronglyEntanglingLayers(params["weights"], wires=range(3), imprimitive=qml.CZ)
        return qml.expval(qml.PauliZ(0) @ qml.PauliZ(1))

    return catalyst.grad(circuit)(params, x)

>>> x = jnp.array([0.1, 0.2, 0.3])
>>> params = {"weights": jnp.ones([3, 3, 3])}
>>> f(params, x)
File catalyst/device/verification.py:238, in validate_observables_parameter_shift.<locals>._obs_checker(obs)
    237 def _obs_checker(obs):
--> 238     if obs and obs.grad_method not in {"A", None}:
    239         raise DifferentiableCompileError(
    240             f"{obs.name} does not support analytic differentiation"
    241         )

AttributeError: 'Prod' object has no attribute 'grad_method'

[dime10]

@erick-xanadu Can check this is resolved?

[erick-xanadu]

While this is no longer raising an exception/error, the test reveals some difference in how qml.grad and catalyst.grad work. I'll put this on hold and look at the other ones for the time being.