Handle zero-variance covariance dimensions in sampling

[sylvesterkaczmarek]

Handle zero marginal variances when normalizing covariance matrices in sample_points() and sigma_points(). Valid positive-semidefinite covariances can have deterministic coordinates, but the current normalization divides their zero rows/columns by zero, producing NaNs and LinAlgError: SVD did not converge.

Use a unit denominator for those zero scale products while retaining the original zero marginal factors. Fixed coordinates therefore remain exactly at their means. Positive-variance behavior, sigma-point scaling, the sqrt=True path, and caller-owned covariance arrays are unchanged.

Validation

Python 3.12, NumPy 2.5.3, SciPy 1.18.1 on macOS arm64:

• New tests on unchanged main (8bb784c): 10 failed, 2 passed. With the fix: all 12 passed.
• Cases cover zero variance in each coordinate, correlated nonzero dimensions, all-zero covariance, singular positive-diagonal controls, explicit fixed dimensions, and a linear unscented transform.
• Deterministic draws verify the resulting means/covariances without statistical tolerances. Divide-by-zero and invalid warnings are promoted to errors.
• Focused utility tests: 42 passed.
• Full suite: 340 passed, 2 skipped, 35 subtests passed.
• New tests pass Flake8; no additional Flake8 findings in utils.py. Syntax checks and git diff --check pass.

No native code or public API changes. Other operating systems were not tested locally.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.39%. Comparing base (45730dc) to head (fa926f0).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #147   +/-   ##
=======================================
  Coverage   97.39%   97.39%           
=======================================
  Files          17       17           
  Lines        5897     5899    +2     
=======================================
+ Hits         5743     5745    +2     
  Misses        154      154           

Flag	Coverage Δ
unittests	97.39% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
ssapy/utils.py	99.74% <100.00%> (+<0.01%)	⬆️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[sylvesterkaczmarek]

Rebased onto current main today; the branch is now 0 behind, mergeable, and the fresh SSAPy matrix is green. The rewritten head is GitHub Verified. Could a maintainer review it when convenient?