docs: fix the README "Result Object Shape" example to match real engine output

[dmchaledev]

Problem

The "Result Object Shape" example in README.md (the npm landing page for this package) is fabricated, internally inconsistent, and missing real fields. For a calculator whose entire value proposition is the numbers it produces, a self-contradictory sample output on the front page is a credibility problem.

The example is implicitly the canonical default input — storage_gb: 70 only occurs for 1000 hosts / medium / weekly / hailbytes_asm / 8h window — yet almost none of the documented numbers match what calculate() actually returns for that input:

Field	README (old)	Real output
cpu_cores	8	6
ram_gb	16	11
ram_recommended	24	17
network_bandwidth_mbps	14	10
optimized_scan_time_minutes	19	250
parallel_hosts	800	600
scan_window_utilization	3.9	52.1
infrastructure_monthly_aws	374	194
infrastructure_monthly_azure	352	184
total_monthly_aws	824	644
total_monthly_azure	802	634
roi.self_managed_monthly	824	644
roi.monthly_savings	525	345
roi.annual_savings	6300	4140
roi.roi_percentage	176.3	115.4

The old numbers don't even agree with each other: parallel_hosts: 800 implies cpu_cores: 8 (consistent), but then optimized_scan_time_minutes should be ceil(1500 / (800/100)) = 188, not the documented 19. And scan_window_utilization: 3.9% is impossible alongside those values for any 1–24h window.

The example also omitted real fields that the engine and index.d.ts define: vm_resources.scaling_info, roi_analysis.self_managed_annual, roi_analysis.managed_annual, and roi_analysis.setup_cost.

Fix

• Replace the example with the verified output of calculate() for the input it already implied, and state that input explicitly above the block.
• Include the previously-missing fields so the documented shape matches the actual runtime/index.d.ts shape.
• Add a regression test (test/calculate.test.mjs → describe('Documentation')) that asserts the documented numbers stay in sync with calculate(), so the example can't silently drift again.

Verification

All numbers were produced by running calculate() against the documented input. npm test passes (66 tests, +1 new).

This is independent of the other open PRs/issues, which focus on calculate() input handling, the time_pressure type, DOM-free import, the test glob, CDN pinning, a11y, and traditional storage sizing — none touch the README result example.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GszLaxg5to1DvSn8cvFiBD

---

Generated by Claude Code