Add weight field to Observable for weighted Pareto analysis

[jc-macdonald]

Summary

Add an optional weight: float = 1.0 field to Observable so users can express relative importance of objectives in Pareto analysis.

Motivation

Currently all observables are treated equally in non-dominated sorting. Users often care more about some objectives than others (e.g. MAE matters more than coverage). Weighted Pareto scales the objective space before NDS, preserving multi-objective structure while reflecting preferences.

Scope

• Add weight: float = 1.0 to Observable dataclass
• Propagate weights into extract_front(), pareto_rank(), hypervolume(), igd_plus()
• Propagate through top_k_pareto_filter(), Study.front(), Study.front_hypervolume()
• Propagate into run_adaptive() (scale optuna objective values)
• Tests for weighted vs unweighted behavior
• Backward compatible: default weight=1.0 preserves existing behavior