fix(home): guard and stabilize dashboard number rendering

[yyqdbngt]

Summary

• Consolidate dashboard number rendering into an exported renderCount() helper: null/undefined/NaN become the existing N/A placeholder, and formatting uses a stable en-US locale (same convention as formatNumber in utils/format.ts)
• Apply it to the broker/proxy/TPS-in/TPS-out table columns and the TPS-out stat card detail
• Guard the "million messages" stat detail against a non-finite totalMessagesToday (renders 0 instead of NaN)
• Pass a guaranteed array to MiniBar when a cluster row omits throughput
• Add regression tests: a cluster with null TPS values and a missing throughput series renders without crashing, plus unit tests for renderCount

Why

Dashboard data comes from the backend over JSON, so field shapes are not guaranteed even when the TypeScript type says otherwise. Two crash/leak paths: v.toLocaleString() on the TPS columns throws for a null value taken from the wire, and MiniBar reads data.length, so a cluster row missing throughput crashed the whole dashboard page (no error boundary). The unqualified toLocaleString() calls additionally produced locale-dependent separators that varied with the viewer's browser locale, inconsistent with the rest of the app's stable formatting.

Testing

• ./node_modules/.bin/vitest run src/pages/home/ → 12 passed (2 new)
• ./node_modules/.bin/tsc --noEmit → clean
• ./node_modules/.bin/eslint src/pages/home/dashboard.tsx src/pages/home/__tests__/DashboardPage.test.tsx → 0 errors

[RockteMQ-AI]

Summary

This PR contains moderate changes (135 lines). I have performed an initial structural review.

Observations

• Changes appear well-scoped
• CLA status: unknown

Maintainers: please verify the implementation details match the intended behavior.

---

Automated review by github-manager-bot

[RockteMQ-AI]

Summary

Defensive fix that improves input validation and error handling. Code looks clean and follows existing patterns.

LGTM

---

Automated review by "github-manager-bot"

[RockteMQ-AI]

LGTM — defensive fix improving input validation and error handling.

---

Automated review by "github-manager-bot"