Fix #1096: MinimumDominatingSet/One to MinimumHittingSet

[isPANN]

Summary

Implement the unit-weight Minimum Dominating Set to Minimum Hitting Set reduction by mapping graph vertices to universe elements and closed neighborhoods to target sets.

This is a stacked PR based on 1075-growth-domain at a9067297 as requested.

Fixes #1096

[isPANN]

Implementation Summary

Changes

• Added MinimumDominatingSet<SimpleGraph, One> -> MinimumHittingSet as a witness-preserving reduction using one closed-neighborhood target set per source vertex.
• Registered exact overhead metadata: universe_size = num_vertices and num_sets = num_vertices, with identity solution extraction.
• Added a canonical five-vertex path example and seven focused tests, including exhaustive equivalence over every simple graph and configuration through four vertices.
• Added a fixture-driven paper theorem and tutorial example, supported by the direct Bannach--Tantau STACS 2018 citation.

Verification

• Independent mathematical verification: constructor 92,121 checks; adversary 102,469 checks; 1,100-instance cross-comparison with 0 disagreements.
• make fmt-check, make clippy, make test, and make paper pass.
• Coverage: 96.19% lines, 95.43% functions, 95.89% regions.

Deviations from Plan

• The pipeline's model guard parsed MinimumDominatingSet/One as a literal model name and reported a false negative. The catalog and source declaration confirm MinimumDominatingSet<SimpleGraph, One> exists, and the issue's completed quality check independently confirms both endpoints.
• Both optimization models are feasible for every instance, so mathematical verification used an infeasible candidate witness instead of a nonexistent infeasible problem instance.
• This base commit does not provide the newer make regenerate-fixtures target; the equivalent cargo run --features example-db --example export_examples command generated the fixture.
• Per maintainer direction, this is a stacked PR based on 1075-growth-domain at a9067297.

Open Questions

• The PR should remain based on 1075-growth-domain until that dependency is merged or otherwise integrated.

[isPANN]

Review pipeline report

Reviewed the PR-specific five-file delta (a9067297..177b561fe462a219fc7aa1c87708408eb8d9f6b8) rather than inherited changes from the stacked base.

Structural and mathematical review

• 12/12 structural checks passed. The rule implementation, ReductionResult, ReduceTo, registration, canonical rule example, tests, and paper entry are all present.
• 6/6 issue-compliance checks passed. The implementation maps vertices to the hitting-set universe, closed neighborhoods to sets, and extracts the same indicator vector.
• The equivalence is sound: a selected vertex set dominates every vertex iff it intersects every closed neighborhood. Feasibility and cardinality are preserved.
• Overhead metadata is exact: universe_size = num_vertices and num_sets = num_vertices.
• Empty, isolated, disconnected, star, complete, and exhaustive simple graphs through four vertices are covered.
• No blacklisted generated files occur in the scoped delta.

Quality review

• DRY, KISS, cohesion/coupling, module placement, and test quality: OK.
• Tests assert the exact transformed structure, identity extraction, boundary cases, and exhaustive semantic equivalence rather than only construction success.
• No critical, important, or minor code-quality findings.

Agentic feature review

The intended unit-weight CLI path passes end to end:

• exact variant and direct reduction discovery: PASS
• canonical rule example (pred create --example MinimumDominatingSet/SimpleGraph/One --to MinimumHittingSet): PASS
• source solve, direct reduction, target solve, source extraction, and evaluation: PASS
• semantic witness check on the five-vertex path: PASS; optimum Min(2) is preserved
• repository cleanliness: PASS

Important finding

1. [Medium] pred inspect advertises an unavailable reduction for the default weighted variant.

   After this unit-weight edge is registered, creating the default MinimumDominatingSet example yields a SimpleGraph/i32 instance. pred inspect lists MinimumHittingSet in reduces_to, but attempting that advertised operation fails:

   $ pred reduce default.json --to MinimumHittingSet -o bundle.json
   Error: No witness-capable reduction path from MinimumDominatingSet to MinimumHittingSet

   This contradiction does not appear at the base SHA because the new edge is absent there. The underlying cause appears to be name-level aggregation in inspect, which includes an edge belonging only to MinimumDominatingSet/SimpleGraph/One while inspecting an i32 instance. Exact unit-weight discovery and reduction work correctly.

The unavailable standalone model-example command for the exact One variant was also checked against the base SHA and is pre-existing; the documented canonical rule example form with --to MinimumHittingSet is added by this PR and works.

Verification

• make test: PASS — 5,438 library tests, 75 integration tests, 161 CLI unit tests, 330 CLI integration tests, 17 pred-sym tests, 26 macro tests, 151 library doctests, and 2 macro doctests.
• make clippy: PASS — all targets with ilp-highs, warnings denied.

Overall: the reduction itself is structurally complete, mathematically correct, well tested, and issue-compliant. One medium CLI discovery/metadata inconsistency should be addressed before merge.