feat: add FinOps model choice opportunity scoring

[ccf]

Summary

• score workflow-level model downshift opportunities inside FinOps cost modeling
• surface potential savings and confidence in the FinOps modeling tab
• true up the roadmap status for model-choice opportunity scoring

Testing

• PYTHONPATH=src pytest tests/test_finops.py -q
• ruff check src/primer/common/schemas.py src/primer/server/services/finops_service.py tests/test_finops.py
• cd frontend && ./node_modules/.bin/vitest run --run src/components/finops/tests/workflow-cost-table.test.tsx src/components/finops/tests/workflow-compare-card.test.tsx src/components/finops/tests/model-choice-opportunity-table.test.tsx
• cd frontend && ./node_modules/.bin/eslint src/components/finops/model-choice-opportunity-table.tsx src/components/finops/modeling-tab.tsx src/components/finops/tests/model-choice-opportunity-table.test.tsx src/types/api.ts
• cd frontend && ./node_modules/.bin/tsc --noEmit

---

Note

Medium Risk
Adds new workflow/model scoring logic to the cost-modeling backend response and surfaces it in the FinOps UI; incorrect heuristics or missing facet/workflow data could yield misleading savings recommendations.

Overview
Extends FinOps cost modeling to compute workflow-level model downshift opportunities by comparing per-archetype success rates and average session costs across models, returning a ranked list plus total_model_choice_savings_monthly in CostModelingResponse.

Adds a new ModelChoiceOpportunityTable card to the Modeling tab to display opportunities, confidence, rationale, and projected monthly savings (with an empty-state), and updates API types/schemas and tests (backend + frontend). Also marks the roadmap item as shipped.

^{Written by Cursor Bugbot for commit 7de10e1. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is ON, but it could not run because the spend limit has been reached. To enable Bugbot Autofix, raise your spend limit in the Cursor dashboard.}

[cursor]

Confidence sort order is inverted, preferring low confidence

Medium Severity

The confidence tiebreaker in opportunities.sort() ranks low-confidence opportunities before high-confidence ones. The mapping {"high": 2, "medium": 1, "low": 0} is sorted ascending by default, so low (0) comes first. The value needs to be negated (like the savings field) so that high-confidence opportunities are preferred when monthly savings are equal. This can cause the [:limit] truncation to drop high-confidence opportunities in favor of low-confidence ones.