v2.5.0

What's new

Optimization API

• — finds the assignment with minimum objective value using branch-and-bound pruning
• — returns a lazy stream of improving assignments; the last element is the global optimum
• — wraps BacktrackingSearch with incumbent-based pruning; sits at the bottom of the solver chain so all preprocessing (node/arc consistency, decomposition) runs first
• Objective functions are called on partial assignments during search — use assignment.getValue(v).orElse(neutralValue) to provide a valid lower bound

Internal

• SolverDecorator base class extracts the common inner/preprocess pattern shared by all decorator solvers, replacing ad-hoc @Value + Solver field boilerplate with a @SuperBuilder hierarchy
• GPG signing moved to the deploy phase so mvn verify no longer prompts for a passphrase