Investigate why some Python expectation value tests failed with NumPy 2

[mhucka]

Describe the issue

When running the Python test cases (in /qsimcirq_tests) with NumPy 2, there are 2 cases that fail unless we loosen the numerical comparison tolerance by 1 order of magnitude. (Specifically, a tolerance of 1e-6 succeeds with NumPy 1, but fails in NP 2 unless I loosen the tolerance to 1e-5.) Those cases are not the only place where a tolerance of 1e-5 shows up in qsimcirq_tests; there's one other in that file. The exact cause is not clear. The failures occur only for the "noisy" versions of the two tests in question, which suggests the cause has something to do with that.

We should investigate this after the next qsim release.

What version of qsim are you using?

0.22.0.dev0

How can the issue be reproduced?

No response