Implement Pareto module wrapping Metaheuristics.jl

[jc-macdonald]

Summary

Port pareto.py functionality using Metaheuristics.jl (JOSS published, NSGA-II/III, HV, IGD+, non-dominated sorting).

Functions to Implement

Function	Description	Metaheuristics.jl API
extract_front(costs, directions)	Extract Pareto-optimal rows	non_dominated_sort
pareto_rank(costs, directions)	Assign Pareto rank to each row	non_dominated_sort
hypervolume(front, ref)	Hypervolume indicator	HV indicator
igd_plus(front, reference_front)	IGD+ indicator	IGDPlus indicator

Design Notes

• Direction flipping (MAXIMIZE → negate before sorting) should mirror the Python implementation.
• Metaheuristics.jl provides all needed indicators natively: GD, IGD, IGD+, HV, spacing, Δₚ.
• Lazy-load via package extension if Metaheuristics.jl is heavy.

Dependencies

[deps]
Metaheuristics = "..."

[compat]
Metaheuristics = "3"

Acceptance Criteria

• All four functions exported and documented
• Direction handling tested (minimize + maximize mixed)
• Property tests: front ⊆ input rows, ranks ≥ 1