Fix test flake in f64::symmetric_eigen by slightly increasing epsilon

[jorgeandrecastro]

Hello!

I saw issue #1361 regarding the test flakes on symmetric_eigen and wanted to try contributing a fix.

What was done: I slightly increased the tolerance epsilon (from 1.0e-5 to 1.0e-4) in the proptests for symmetric_eigen and symmetric_eigen_singular to prevent false positives caused by floating-point rounding variations on certain random matrices of size 14.

Testing: I ran the tests locally (cargo test symmetric_eigen), and everything passed successfully.

I am thrilled to try contributing to this awesome project! Please let me know if this approach is okay, if there are better practices to follow, or if you disagree with anything. I am entirely open to feedback.

Thank you very much for your guidance!