feat(simulation): introduce unit-consistent objective semantics

[div0rce]

Summary

• Introduces objectiveUtilityCents and utility_usd_cents as the explicit live ranking unit while keeping legacy score as a compatibility alias.
• Converts cashback and issuer points through documented objective utilities, adds interpreted score components, and labels liquidity pressure as bounded heuristic contribution.
• Adds objective semantics docs, a narrow guardrail for mixed-unit objective math, regression tests, and engine behavior v6 replay coverage.

This PR does not make Cherry globally optimal. It makes live candidate ranking unit-consistent, interpretable, and guardrailed against mixed-unit scoring.

Testing

• CHERRY_TMP_ROOT="$HOME/.cherry-tmp" npm run lint
• CHERRY_TMP_ROOT="$HOME/.cherry-tmp" npm run check
• npm run typecheck
• npm run typecheck:scripts
• CHERRY_TMP_ROOT="$HOME/.cherry-tmp" npm test
• CHERRY_VINE_SIGNATURE_MODE=enforce npm run build
• CHERRY_TMP_ROOT="$HOME/.cherry-tmp" CHERRY_VINE_SIGNATURE_MODE=enforce npm run verify:repo-closure

Risk

• Changed domains: engine objective scoring, solver payload shape, objective docs, guardrail registry, replay behavior version.
• Compatibility: existing score remains present as an alias for objectiveUtilityCents; legacy components remain present alongside scoreComponents.
• Possible regressions: consumers that assume only the old decision fields may see additional serialized fields; ranking can change where issuer points now have explicit objective value.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
cherry	Ready	Preview, Comment	Apr 28, 2026 10:02pm