Fix CI regressions after Pareto search merge - #1124

Why CI failed

The Pareto refactor left one stale single-path variable in the example-db test, while the example and differential fixture also triggered Clippy errors under the CI toolchain.

Fix

• inspect only the authored direct witness edge for each example-db rule; indirect paths are not substitutes for that rule’s stored solution pairs
• keep aggregate/proof-only capability checks explicit when no direct witness edge exists
• remove redundant references in the chained reduction example
• update the small differential oracle fixture for current Clippy rules

This does not restore a winner-selection API, a best-path recommendation, or a compatibility fallback.

Verification

• cargo fmt --all -- --check
• CARGO_BUILD_JOBS=1 cargo check --workspace --tests --features "ilp-highs example-db"
• CARGO_BUILD_JOBS=1 cargo clippy --all-targets --features ilp-highs -- -D warnings
• CARGO_BUILD_JOBS=1 cargo test -p problemreductions --lib --features "ilp-highs example-db" rule_specs_solution_pairs_are_consistent -- --test-threads=1
• CARGO_BUILD_JOBS=1 cargo test -p problemreductions --lib --features ilp-highs test_pareto_search_matches_independent_small_graph_oracle -- --test-threads=1

Scope

Three files only. No resource model, search contract, CLI surface, or public API changes.