FaultEnumerator with method='propagator' incorrectly handles multiple classical registers

[quantumjim]

Steps to reproduce the problem

from qiskit_qec.analysis.faultenumerator import FaultEnumerator
from qiskit import *

qc = QuantumCircuit(2)
for _ in range(2):
    qc.add_register(ClassicalRegister(1))
qc.measure(range(2),range(2))

fe = FaultEnumerator(qc, order=1, method="propagator")

print(list(fe.generate()))

What is the current behavior?

Output is flattened to a single classical register when FaultEnumerator is used with method="propagator". For example, the following outputs

[(0, ['measure'], ['x'], [1, ' ', 0]), (1, ['measure'], ['x'], [0, ' ', 1])]

for method="stabilizer", where the [1, ' ', 0] represents the bit string 1 0 that would be output by the circuit, etc. However, for method="propagator" the output is

[(0, ['measure'], ['x'], [1, 0]), (1, ['measure'], ['x'], [0, 1])]

What is the expected behavior?

The two methods should give the same output.

This issue has been fixed in #179 for the use of generate(), but the fix has not yet been applied to generate_blocks().

Once the issue is fixed, the following test should be added to 'tests/matching/test_repetitionmatcher.py' (added in #179) and should pass.

    def test_error_pairs_pymatching_propagator(self):
        """Test the case with two faults on a d=5 code using pymatching and the propagator method."""
        self.test_error_pairs(dec_method="pymatching", fe_method="propagator")