Unify reduction construction, result recovery, and solver boundaries

[isPANN]

This PR makes an executed reduction responsible for constructing its target and recovering the complete source result. Solvers and CLI workflows use that same recovery path, including when a target optimum establishes source infeasibility.

API and behavior

Before	After
Solution extraction, value extraction, and optional completion callbacks carried separate parts of recovery.	ReductionResult::recover_result(source, target_result) returns the complete source result.
Witness and aggregate execution could construct the same reduction separately.	Each executed step owns one result shared through Rc; shared path prefixes reuse that execution.
A target configuration did not carry a uniform statement of solution quality.	SolveOutcome distinguishes Optimal, Feasible, and Infeasible; execution failures remain errors.
Decision bounds and optimization targets could require separate interpretation during completion.	Decision targets use Decision<P> where appropriate; rule-owned recovery handles bounds and penalty semantics.
External extraction accepted a raw configuration.	pred extract bundle.json --result target-result.json accepts a complete result, validates the target witness, and follows the same recovery path as bundle solving.

For example, a penalty QUBO can have an optimum even when its source ILP is infeasible. The ILP-to-QUBO rule interprets its established energy relation and returns source Infeasible. A merely feasible candidate that cannot establish the source result returns InsufficientSolutionQuality, never a proof of infeasibility.

let reduction = source.reduce_to()?;
let target_result = SolveOutcome::optimal(reduction.target_problem(), target_solution)?;
let source_result = reduction.recover_result(&source, target_result)?;

The solver establishes optimality; SolveOutcome::optimal packages that result and evaluates the witness. Each rule implements its recovery semantics explicitly. There is no default recovery strategy or new cross-rule recovery abstraction.

Supporting changes

• Keep brute-force candidate aggregation within the brute-force solver, and use registered solver capabilities for dispatch.
• Store QUBO matrices using sprs, reuse ILP buffers, and use existing petgraph algorithms for applicable graph operations.
• Validate deserialized models and reconstruct derived state through model construction paths.
• Correct equal-size pairing recovery in the 3-dimensional-matching-to-3-partition rule.
• Remove the unused ProblemMetadata trait and point documentation to catalog lookup; narrow internal helper visibility.
• Migrate callers, examples, documentation, and existing regression cases to the new contract. Preserve the existing test cases, adjusting their API usage.

This is a package-wide API migration. The broad file count includes model/rule callers and their tests. Verifier tooling, local skills, and generated reports are excluded.

Validation

• make check
• cargo clippy --workspace --all-targets --all-features -- -D warnings
• Focused registry, CLI, QUBO recovery, and documentation tests
• Earlier full coverage run: 97.36% relative to origin/main (before the final metadata cleanup)
• Earlier MCP tests and paper build passed

Refs #1148.

[codecov]

Codecov Report

❌ Patch coverage is 95.24002% with 236 lines in your changes missing coverage. Please review.
✅ Project coverage is 96.42%. Comparing base (7dd5fcd) to head (e99c577).

Files with missing lines	Patch %	Lines
...c/models/graph/shortest_weight_constrained_path.rs	81.94%	13 Missing ⚠️
src/models/misc/minimum_weight_and_or_graph.rs	79.31%	12 Missing ⚠️
src/models/graph/undirected_flow_lower_bounds.rs	76.74%	10 Missing ⚠️
...els/graph/hamiltonian_path_between_two_vertices.rs	73.52%	9 Missing ⚠️
src/models/misc/flow_shop_scheduling.rs	72.72%	9 Missing ⚠️
src/models/misc/job_shop_scheduling.rs	79.06%	9 Missing ⚠️
...rc/models/misc/minimum_fault_detection_test_set.rs	81.25%	9 Missing ⚠️
...c/models/misc/precedence_constrained_scheduling.rs	82.00%	9 Missing ⚠️
...dels/graph/directed_two_commodity_integral_flow.rs	87.09%	8 Missing ⚠️
src/models/misc/feasible_register_assignment.rs	83.67%	8 Missing ⚠️
... and 53 more

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1151      +/-   ##
==========================================
+ Coverage   95.93%   96.42%   +0.49%     
==========================================
  Files        1074     1070       -4     
  Lines      132106   144318   +12212     
==========================================
+ Hits       126730   139160   +12430     
+ Misses       5376     5158     -218     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.