Strange behavior of annealing time

[maqid]

Describe the bug
When the same problem is annealed by different num_sweeps, even though num_sweep is set smaller, the annealing time may be long.

import openjij as oj

N = 30
h = {0: -10}
J = {(i, i+1): 1 for i in range(N-1)}

num_reads = 10

for num_sweeps in [25, 50]:
    sampler = oj.SASampler(num_sweeps=num_sweeps, num_reads=num_reads)
    response = sampler.sample_ising(h, J)
    t = response.info["execution_time"]
    print(f"t(num_sweeps={num_sweeps}) = {t}")

t(num_sweeps=25) = 88.32270978018641
t(num_sweeps=50) = 58.8218797929585

Expected behavior

• t(num_sweeps=25) < t(num_sweeps=50)

Environment (please complete the following information):

• OS: ubuntu18.04
• Version python 3.9.6