[Story]: BU rollup data model over per-person budgets

## Summary
Build the BU-rollup data model that aggregates existing per-person µ$ spend and budget data up the organization hierarchy to region/BU/company level.

## Intent / Source of truth
The enforcement data (per-person spend) already exists (ai-helm ADR-0021/0035); this story is purely the aggregation layer, and depends on the organization hierarchy epic's data model. Part of [Epic] Per-BU cost attribution and chargeback reporting.

## Scope
- [ ] Rollup aggregation query/materialized-view over per-person spend, keyed by org-tree node
- [ ] Reconciliation check: rollup total always equals the sum of underlying per-person records
- [ ] API to fetch spend at any org-tree node

## Out of scope
- Export formatting (separate story)
- Real-time alerting (separate story)

## Verification
Automated reconciliation test comparing rollup totals against raw per-person ledger sums across the org tree, run on a realistic dataset.

## Risk assessment
Depends on the organization hierarchy epic (a2) shipping its data model first; sequence this story after that epic's a2s1/a2s4.

## AI Usage Declaration
Drafted with AI assistance during the 2026-08-13 cross-repo backlog consolidation and enterprise-readiness research. A human owns intent, verification and consequences.

Start with the organization hierarchy epic, especially a2s1/a2s4, and read ai-helm ADR-0021/0035 to understand the existing per-person enforcement data. Once that data model is available, define the rollup query or materialized view and API scope, then add the automated reconciliation test over a realistic dataset; done means org-tree totals match the raw per-person ledger sums.

由索引模型根据 Issue 内容生成。