Add monitoring center for API usage metrics

[liuw1535]

Motivation

• Provide a lightweight monitoring center to show API call counts and token usage for demo/system presentation, with selectable ranges (7/14/30 days), summary cards (requests, tokens, TPM, RPM, RDP) and a token-ranked per-model donut chart.

Description

• Add a persisted usage statistics store src/utils/usageStats.js that keeps daily buckets (up to 30 days) and exposes usageStats and setUsageMetrics for recording and summarizing metrics; data persisted to data/usage-stats.json.
• Insert a request-monitoring middleware in src/server/index.js that infers model names for common API paths and records metrics on response finish (only tracks /v1/, /v1beta/, /cli/v1/).
• Wire usage recording into response paths by calling setUsageMetrics from handlers src/server/handlers/openai.js, src/server/handlers/gemini.js, src/server/handlers/claude.js, and src/server/handlers/geminicli.js for stream/fake/non-stream branches so usage/status are captured.
• Add admin API GET /admin/monitor/usage in src/routes/admin.js that returns usageStats.getSummary(days) for the frontend.
• Implement frontend Monitor tab: public/index.html (tab and page), new UI logic in public/js/monitor.js, tab integration in public/js/ui.js and public/js/main.js, and styles in public/css/pages.css to render the five summary cards and a conic-gradient donut chart with model token distribution.

Testing

• Ran syntax checks: node --check on modified server and frontend files (all succeeded).
• Performed a smoke test using the bundled usage API: imported usageStats and executed recording calls then validated usageStats.getSummary(7) contains the recorded requests and model distribution (script run succeeded and printed usage stats smoke test passed).
• Verified basic integration by running the linter/quick checks used above and ensuring no runtime parse errors across changed files.

---

Codex Task