Memory leak when using numpy matrix in tests

Hi, when using non-empty numpy matrices in tests they will not be garbage collected and pile up slowly using more and more memory.

pip list output:

attrs          22.2.0 
exceptiongroup 1.1.1  
iniconfig      2.0.0  
numpy          1.24.2 
packaging      23.0   
pip            10.0.1 
pluggy         1.0.0  
pytest         7.2.2  
setuptools     39.1.0 
tomli          2.0.1 

Operating system:
Ubuntu 18.04.6 LTS
python 3.8

Minimal example:

import numpy as np


class TestMemory:
    def test_memory(self):
        for _ in range(1000000):
            test = np.matrix(np.random.rand(100, 100))

np.array does not share the same issue, so one can avoid matrix as much as possible, but for some operations it is still needed.

Start by running the minimal TestMemory example with the listed Python, pytest, and numpy versions, then compare its memory behavior with the np.array variant. Investigate why non-empty numpy.matrix values are retained during tests. Done means repeated matrix creation no longer causes memory usage to accumulate, while the relevant test behavior remains intact.