feat(expenditures): count only approved expenditures in totals

[nourshoreibah]

Problem

Nothing that reports money spent filtered on expenditures.status. Every SUM(amount), chart series, budget percentage and report total counted pending, denied and needs_more_info rows alongside approved ones — so an expenditure inflated the spend figures the moment someone entered it, before anyone reviewed it, and a project could render as over budget on the strength of requests that were later denied.

The only status-aware code was the write path (PATCH /expenditures/{id}/status), the status badge, and the optional user-selected filter on the expenses list page.

Change

Filter to status = 'approved' everywhere a figure is computed:

projects lambda

• GET /dashboard — totalSpent, topExpenseCategory, the averageSpendPerProject numerator, the per-project spent on the cards, and the expensesByMonth series behind the bar chart
• GET /projects — total_spent on the list cards (loadProjectAggregates)
• GET /projects/{id}/overview — totalSpent, totalRemaining, spentPercentage, expenditureCount

reports lambda

• fetchReportData — the expenditures the PDF and DOCX tables list, and therefore the total each prints

Deliberately left unfiltered

The raw lists — the expenses table on the project detail page, GET /projects/{id}/expenditures, and the /expenditures admin review queue. Their job is to show what is awaiting review. On the overview endpoint the table still returns every row while the stats beside it count only approved ones.

No frontend change was needed: the only client-side reduction (ExpensesBarChart) sums the already-filtered series the dashboard returns.

Seed

Seeded expenditures had no explicit status, so they all defaulted to pending — after this change the local dashboard would have read $0. The seed now sets statuses explicitly and keeps one denied and one pending row, which gives the admin review queue something to show and keeps the seed honest about what the totals exclude. Row count and amounts are unchanged; seed spend is 18000, of which 14000 is approved.

Tests

• projects/test/approved-expenditures.e2e.test.ts (new, 8 tests) — one project, one expenditure per status; asserts every dashboard figure, the list card total and the overview stats see only the approved 1000, and that the overview table still returns all four rows.
• reports/test/report-service.e2e.test.ts (new, 2 tests) — fetchReportData returns approved rows only, and the total the report prints follows.
• projects/test/projects.e2e.test.ts — the one existing assertion that moved: seed totalSpent 18000 → 14000, average 4500 → 3500. topExpenseCategory (Travel/6800) and project 1's 9200 are unchanged by design, so they still pin the non-status behaviour.
• expenditures/test/expenditures.e2e.test.ts — three PATCH /expenditures/{id}/status rejection tests asserted the row was still 'pending' afterwards, which encoded the seeded status rather than what they mean: that the write did not happen. They now capture the status before the call and compare, and ask for a status the row does not already hold, so a rejection test cannot pass merely because the request matched what was already there. The approve and decline cases move to the seeded pending row so they still exercise a real transition.

reports/jest.config.js gains maxWorkers: 1: it now has two suites that reseed the same database, which truncate each other's fixtures when run in parallel. The projects lambda already does this for the same reason.

Verification: full CI green — every lambda suite, frontend-ci, migrations-fresh and shared-auth.

🤖 Generated with Claude Code