Add perturb_fraction / random_state to make_neighbor_perturbation

[dbdimitrov]

Summary

• make_neighbor_perturbation (and _make_perturbed_expression / _node_perturbation) gain perturb_fraction=1.0 and random_state=0.
• Only a seeded random subset of round(perturb_fraction * n_cells) cells keeps the shift; the rest are reverted to their original expression after clipping and before renormalisation. Out-of-range values raise ValueError.
• Defaults reproduce the previous behaviour exactly.
• New test test_node_perturbation_fraction in tests/test_cellina.py (fraction 1.0 == old, 0.0 == unchanged, 0.5 changes half the rows deterministically, invalid range raises).

Distilled from 0f5b68a on the rebuttal branch (only the src/cellina/_spatial_utils.py hunks; the analysis/ scripts are not included). Required by the node-fraction dose-response sweep in cellina-reproducibility PR #59.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.17%. Comparing base (c77b614) to head (24fe743).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #39      +/-   ##
==========================================
+ Coverage   90.10%   90.17%   +0.07%     
==========================================
  Files          10       10              
  Lines        1394     1405      +11     
==========================================
+ Hits         1256     1267      +11     
  Misses        138      138              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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