fix: maintain valid state through the algorithm

[john-halloran]

Previously the loose usage of "self.components_" was causing candidate components to be mixed with stored ones during a component update. This fix better matches MATLAB behavior and improves recovery, especially with a broader range of datasets that I previously hadn't tested with.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.32%. Comparing base (d7586bc) to head (a0e744a).
⚠️ Report is 14 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #210      +/-   ##
==========================================
+ Coverage   77.50%   87.32%   +9.82%     
==========================================
  Files           3        3              
  Lines          40       71      +31     
==========================================
+ Hits           31       62      +31     
  Misses          9        9              

Files with missing lines	Coverage Δ
tests/test_snmf_optimizer.py	100.00% <100.00%> (ø)

🚀 New features to boost your workflow:

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

[sbillinge]

This probably could use a test for the thing that broke that allowed you to find it in the first place, if you can engineer that. It would be a test that failed with the old code but passes with your current code fix.

[john-halloran]

This probably could use a test for the thing that broke that allowed you to find it in the first place, if you can engineer that. It would be a test that failed with the old code but passes with your current code fix.

Okay, they've been added for both the cubic root solver and the component update loop.

[sbillinge]

Brilliant! Thanks so much.