TST: eval test unreliable

[rhshadrach]

Ran into an unrelated failure I haven't seen before in #37227. Occurs in pandas/tests/computation/test_eval.py:211

https://dev.azure.com/pandas-dev/pandas/_build/results?buildId=45754&view=logs&j=89027c74-d1cd-5e5d-942f-e6a83ebe07fe&t=768a7acf-33bc-5b3d-16e1-da874611990b&l=92

E           numpy array values are different (2.0 %)
E           [left]:  [[10.885282508088554, 24.01653292315993, 2.5334764666276754, -26.20850153209755, -9.292195139788854], [4.631505078070531, 0.43305142455418066, 9.796254287558813, -0.817344630073512, -0.22979400652647], [16.964677878827604, 3.69159983714683, 0.16736723603895715, -13.866779975183137, -0.6547978574323233], [0.9332007475405953, 7.326122317479498, 2.2912333245159338, -0.9957313583066364, -0.21968895568980523], [1.2901449066742292, 1.5178122458579082, 0.3101899430985658, -0.3600640963287845, -3.8622009358387115], [1.7324531378555383, 1.6717678316305915, 0.39248288379342006, -2.259323953594114, -13.215192982180358], [60.71781357885021, 13.08256705151808, 3.306197592158367, -2.603163236030203, -0.5783506355776388], [47.71284995565457, 39.508260846448245, 0.16695497225705538, -7.39400467084908, -0.9437156402277913], [0.40861584564047, 611.3601636482858, 0.8350779591410551, -3.034557944041943, -0.33953697203938915], [46.62807235120426, 6.308023182021756, 5.01671850424403, -0.23877079736974396, -0.5460348132251863]]
E           [right]: [[10.885282508088554, 24.01653292315993, 2.5334764666276754, -26.20850153209755, -9.292195139788854], [4.631505078070531, 0.43305142455418066, 9.796254287558813, -0.817344630073512, -0.22979400652647], [16.964677878827608, 3.69159983714683, 0.16736723603895715, -13.866779975183137, -0.6547978574323233], [0.9332007475405953, 7.326122317479498, 2.2912333245159338, -0.9957313583066364, -0.21968895568980523], [1.2901449066742292, 1.5178122458579082, 0.3101899430985658, -0.3600640963287845, -3.8622009358387115], [1.7324531378555383, 1.6717678316305915, 0.39248288379342006, -2.259323953594114, -13.215192982180358], [60.71781357885021, 13.08256705151808, 3.306197592158367, -2.603163236030203, -0.5783506355776388], [47.71284995565457, 39.508260846448245, 0.16695497225705538, -7.39400467084908, -0.9437156402277913], [0.40861584564047, 611.3601636482858, 0.8350779591410551, -3.034557944041943, -0.33953697203938915], [46.62807235120426, 6.308023182021756, 5.01671850424403, -0.23877079736974396, -0.5460348132251863]]

where one entry is 16.964677878827604 and the other is 16.964677878827608. Haven't looked into it, but guessing this is a numerical precision issue.

[avinashpancham]

take