Add normalized CheapJPDAF with linear-cost soft association

[FlorianPfaff]

Summary

Implement CheapJointProbabilisticDataAssociationFilter, exported as CheapJPDAF and CJPDAF, for NumPy linear-Gaussian multitarget tracking.

• Replace combinatorial joint-event enumeration with Fitzgerald-style row/column competition in O(n_targets * n_meas) association time and memory.
• Reuse JPDAF's gating, Gaussian likelihoods, per-pair Kalman hypotheses, prediction, and moment-matched state/covariance updates. The existing exact event solver is extracted into an overridable method without changing its mathematics.
• Compute association odds and exclusive competing-track sums in log space, avoiding cancellation and global rescaling across disconnected components.
• Add public exports, experimental API registry entries, backend capability metadata, generated documentation tables, and an executable usage guide in docs/cheap-jpdaf.md.

Normalization convention

For gated likelihoods L and clutter intensity kappa, define w_ij = P_D L_ij / ((1-P_D) kappa_j), r_i = sum_j w_ij, and c_j = sum_i w_ij. Detection weights are

beta_ij = w_ij / (1 + r_i + c_j - w_ij)

The missed-detection mass is 1 - sum_j beta_ij, evaluated through an equivalent sum of positive terms to retain tiny miss probabilities. The original separate Fitzgerald miss expression 1/(1+r_i) generally does not normalize under track competition; this deliberate complementary-mass convention is documented rather than silently returning invalid mixture weights.

The approximation has normalized per-track mixtures and measurement-column allocation at most one. It agrees with the current exact JPDAF model for a single track, a single measurement, and disjoint single-track validation components; it is not exact in general ambiguous scenes.

Diagnostic contract

The optional hard diagnostic is a feasible track-order greedy likelihood-ratio assignment, not a joint MAP event. It is returned by find_association and stored as latest_greedy_association; latest_map_association remains None. The diagnostic never drives the soft update. max_enumerated_events is ignored by the cheap subclass and remains enforced by exact JPDAF.

Verified validation

GitHub-hosted Ubuntu / Python 3.12 validation completed successfully:
https://github.com/FlorianPfaff/PyRecEst/actions/runs/34966924288

• 47 passed: new cheap JPDAF regressions, existing exact JPDAF regressions, and the existing likelihood log-determinant underflow regression.
• Includes 100 seeded randomized matrix checks, exact reference special cases, extreme likelihood scales, permutation equivariance, event-limit independence, mixture covariance, empty/gated scans, vector clutter, heterogeneous measurement covariances, and backend/input contracts.
• Pylint: 10.00/10 for the changed tracker modules and new tests.
• Backend matrix and public API registry checks passed.
• Minimal-import smoke test passed.
• Documentation example executed successfully.

The validated integrated source was pushed as cb1abef77903d79c54289ce3ab427c4ccf186f43. Temporary integration workflow/script files were removed from the final branch. The broader repository CI matrix has not yet been confirmed; this PR remains draft pending those checks. No merge or release is performed.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Max errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no		no	6.5s
✅ JSON	prettier	7	0	0		0	0.56s
✅ JSON	v8r	7		0		0	4.29s
✅ MARKDOWN	markdownlint	69	0	0		0	2.17s
✅ MARKDOWN	markdown-table-formatter	69	0	0		0	0.47s
✅ PYTHON	black	2042	1	0		0	53.9s
✅ PYTHON	isort	2042	8	0		0	2.42s
✅ REPOSITORY	betterleaks	yes		no		no	1.26s
✅ REPOSITORY	checkov	yes		no		no	30.99s
✅ REPOSITORY	git_diff	yes		no		no	0.1s
✅ REPOSITORY	secretlint	yes		no		no	5.87s
✅ REPOSITORY	syft	yes		no		no	2.67s
✅ REPOSITORY	trivy-sbom	yes		no		no	2.22s
✅ YAML	prettier	11	0	0		0	0.56s
✅ YAML	v8r	11		0		0	12.15s
✅ YAML	yamllint	11		0		0	0.61s

Notices

⚠️ Your configuration references items that have been removed from MegaLinter and are ignored: MAKEFILE_CHECKMAKE. See Removed linters to find their replacements.

See detailed reports in MegaLinter artifacts

Your project could benefit from a custom flavor, which would allow you to run only the linters you need, and thus improve runtime performances. (Skip this info by defining FLAVOR_SUGGESTIONS: false)

• Documentation: Custom Flavors
• Command: npx mega-linter-runner@10.1.0 --custom-flavor-setup --custom-flavor-linters PYTHON_BLACK,PYTHON_ISORT,COPYPASTE_JSCPD,JSON_V8R,JSON_PRETTIER,MARKDOWN_MARKDOWNLINT,MARKDOWN_MARKDOWN_TABLE_FORMATTER,REPOSITORY_CHECKOV,REPOSITORY_GIT_DIFF,REPOSITORY_BETTERLEAKS,REPOSITORY_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY_SBOM,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository