Running `:func: Pattern.perform_pauli_measurements` twice returns non-runnable patterns.

[matulni]

The transformation $T$ implemented in :func: Pattern.perform_pauli_measurements should fulfill

$$ T^n[P] = T[P], \quad n > 0$$

The example below shows that apply :func: Pattern.perform_pauli_measurements twice returns a non-runnable pattern.

import numpy as np
from graphix.random_objects import rand_circuit
import pytest

rng = np.random.default_rng(42)
nqubits = 2
depth = 2
circuit = rand_circuit(nqubits, depth, rng, use_ccx=False)
p = circuit.transpile().pattern

p.perform_pauli_measurements()
print(p.results)  # {0: 0, 2: 0, 3: 0, 4: 0, 5: 0, 7: 0, 8: 0, 1: 0, 10: 0, 11: 0, 12: 0, 13: 0, 14: 0, 9: 0, 16: 0, 15: 0, 18: 0}
p.perform_pauli_measurements()
print(p.results)  # {}
p.check_runnability() # Raises error

I believe this occurs because :func: Pattern.perform_pauli_measurements does not handle Pattern.results adequately. If so, it should be an easy fix that could be incorporated to #392.

[emlynsg]

I believe I fixed this with the latest commit - I have just added a check to the start of the function which just returns the pattern as-is if pauli preprocessing has been done. However, there might be cases where patterns could be manipulated with intervening steps that effectively make them no longer pauli preprocessed, but for which the Pattern class flag _pauli_preprocessed is not updated. Worth a quick discussion.