``DWaveCliqueSampler()`` initial embedding slower than I expected

[JoelPasvolsky]

Description
This might be a local problem with my system getting old but the initial embedding is taking quite a lot longer than I would have expected, especially on Win OS. In the past it took a couple of minutes, now it is over 5 minutes.
Just raising this in case some change has inadvertently slowed this down.

To Reproduce

Win OS:

>>> timeit.timeit('from dwave.system import DWaveCliqueSampler; qpu = DWaveCliqueSampler(solver="Advantage_system4.1"); print(qpu.child.solver.name); print(qpu.largest_clique_size)', number=1)
Advantage_system4.1
177
Out[18]: 402.3655947998632

>>> timeit.timeit('from dwave.system import DWaveCliqueSampler; qpu = DWaveCliqueSampler(solver="Advantage_system6
    ...: .4"); print(qpu.child.solver.name); print(qpu.largest_clique_size)', number=1)
Advantage_system6.4
175
Out[19]: 51.09168740012683

Unix:

>>> timeit.timeit('from dwave.system import DWaveCliqueSampler; qpu = DWaveCliqueSampler(solver="Advantage_system4.1"); print(qpu.child.solver.name); print(qpu.largest_clique_size)', number=1)
Advantage_system4.1
177
Out[16]: 224.93166744199698

Expected behavior
I remember a couple of minutes and see here 152 seconds on Pegasus (which I am guessing was also Advantage_system4 but am not sure)

Environment:

• OS: Ubuntu 24, and Win 10
• Python version: 3.12

Additional context
Note the large difference between Advantage_system4 and Advantage_system6, perhaps the time is going on the cliques for 176 & 177 nodes?

[arcondello]

Nothing has changed so far as I know, I would recommend A/B testing with an older version of minorminer to get a baseline.

[boothby]

I'd contrast with with 0.2.15; the last time the build system was touched. Also, I can eyeball the expected relative performance of the algorithm if you plot these graphs with dnx.draw_pegasus_yield.