More performance improvements for Collect2qBlocks

[IvanIsCoding]

Summary

Related to Qiskit/rustworkx#265

Uses the ported version of the single-sweep algorithm from #6534 in retworkx.

This PR brings a performance boost and removes the bottleneck from the graph traversal

Details and comments

Depends on retworkx 0.10 being released Retworkx 0.10 has been released, this PR is ready for review

[IvanIsCoding]

Results show roughly a 1.25x speed up compared to the main branch. Using the same benchmark from #6534:

import timeit
from qiskit import QuantumRegister, QuantumCircuit
from qiskit.circuit.library import QuantumVolume
from qiskit.converters import circuit_to_dag
from qiskit.transpiler.passes.optimization import Collect2qBlocks
from qiskit.compiler import transpile

N = 100
q = QuantumRegister(N, 'q')
qc = QuantumCircuit(q)
qc.append(
        QuantumVolume(N, N//2, seed=2, classical_permutation=False), q
)

qc = transpile(qc, optimization_level=0, basis_gates=['u1', 'u2', 'u3', 'cx'])
dag = circuit_to_dag(qc)
c2q = Collect2qBlocks()

print(
        timeit.timeit("c2q.run(dag)", globals=globals(), number=10)
)

Main: 0.89404s
#6680: 0.70438s

[ajavadia]

Are the logic improvements here reflected in the retworkx version?
#6534

[IvanIsCoding]

Are the logic improvements here reflected in the retworkx version?
#6534

Yes, it should be identical. Which means this PR should be on hold until the issues #6686 are explored further

[mtreinish]

LGTM, thanks for doing this.

[mtreinish]

I removed the automerge tag, since as @jakelishman reminded me this needs an upgrade release note documenting the increase in the minimum retworkx version used. I'll add this shortly and just merge it after