Unexpected JIT compilation error when cudaq.observe from Python

[bmhowe23]

Required prerequisites

• Consult the security policy. If reporting a security vulnerability, do not report the bug using this form. Use the process described in the policy to report the issue.
• Make sure you've read the documentation. Your issue may be addressed there.
• Search the issue tracker to verify that this hasn't already been reported. +1 or comment there if it has.
• If possible, make a PR with a failing test to give us a starting point to work on!

Describe the bug

Running cudaq.observe() from a Python test program shown below produces the following error:

error: 'cc.address_of' op must reference a global
RuntimeError: Could not apply measurements to ansatz.

Steps to reproduce the bug

Save the following to test.py

import cudaq
from typing import List

cudaq.set_target("oqc", emulate=True)

# Initialize a kernel/ ansatz and variational parameters.
@cudaq.kernel
def kernel(angles: List[float]):
    # Allocate a qubit that is initialized to the |0> state.
    qubit = cudaq.qubit()
    # Define gates and the qubits they act upon.
    rx(angles[0], qubit)
    ry(angles[1], qubit)


# Our Hamiltonian will be the Z expectation value of our qubit.
hamiltonian = cudaq.spin.z(0)

# Initial gate parameters which initialize the qubit in the zero state
initial_parameters = [0, 0]

cost_values = []

def cost(parameters):
    """Returns the expectation value as our cost."""
    expectation_value = cudaq.observe(kernel, hamiltonian,
                                      parameters).expectation()
    cost_values.append(expectation_value)
    return expectation_value

# We see that the initial value of our cost function is one, demonstrating that our qubit is in the zero state
initial_cost_value = cost(initial_parameters)
print(initial_cost_value)

And then run the test program:

$ python3 test.py
error: 'cc.address_of' op must reference a global
RuntimeError: Could not apply measurements to ansatz.

Expected behavior

The program should run without errors. Note that a very similar program in C++ (targettests/execution/cudaq_observe.cpp in the repo) seems to run without errors.

Is this a regression? If it is, put the last known working version (or commit) here.

Yes; last known to have worked in 0.8.0

Environment

• CUDA-Q version: latest (which is 3300b12 as of the time of this writing)
• Python version: 3.10
• C++ compiler: N/A
• Operating system: N/A

Suggestions

No response