Update Impact Tab to use Cloudflare Edge Data

[isTravis]

Replaces the Metabase/Redshift analytics pipeline with per-community analytics sourced directly from Cloudflare's edge data. Staged as "Impact2" alongside the existing Impact tab so nothing is removed while we test.

What this adds

Server

• server/utils/cloudflareAnalytics.ts: Cloudflare GraphQL Analytics API client. Single combined query fetches daily traffic, top paths, countries, devices, and referrers per hostname. Includes date chunking (CF max ~31 days per query), contiguous span grouping, and per-day breakdown caching.
• server/analyticsDailyCache/model.ts: Sequelize model for Postgres-backed daily cache. Past days are cached permanently; today's data has a 1-hour TTL. Auto-prunes rows older than 90 days (hourly throttle).
• server/impact2/api.ts: API routes (GET /api/impact2, /test, /debug). Resolves the Cloudflare-facing hostname by querying the Community model directly (bypasses getInitialData's localhost domain overwrite in dev). Debug endpoint is disabled in production.
• Registered in server/apiRoutes.ts, server/routes/index.ts, server/models.ts.

Client

• client/containers/DashboardImpact2/: React component + SCSS. Top row with stat cards (color-coded left border) and area chart side by side. Four-column data grid below: Top Pages (with clickable links), Countries, Referrers, Devices (as percentage table). Responsive down to single-column on mobile.
• Date range picker: Today / 7 days / 30 days.
• Graceful degradation: friendly "not available" state when env vars are missing, stale-data callout when CF is unreachable but cache exists, standard error state with retry for transient failures.
• Registered in client/containers/App/paths.ts, client/containers/index.ts, utils/dashboard.ts.
• Nav button added to ScopeDropdown.

Filtering and noise reduction

Two layers of filtering to get numbers closer to real human readership:

1. Cloudflare query filters: Only counts requests that are successful (2xx/3xx), serve HTML content, use GET method, and come from eyeball sources (not cloudflare internal routing).
2. Server-side noise path filter: Removes /wp-*, /cdn-cgi/, /api/, /static/, /login, /robots.txt, .xml, etc. from Top Pages list. Subtracts noise path page views from totals and proportionally scales visits, countries, devices, referrers, and daily chart data to match.

Raw (pre-adjustment) totals are included in the API response and shown in the footer for transparency.

Caching strategy

• 0 Cloudflare API calls when all requested days are cached and within TTL.
• Past days: permanent cache (data is final).
• Today: 1-hour TTL, then re-fetched from CF on next request.
• Cache is Postgres-backed, so it persists across swarm deploys.

Env vars required

CLOUDFLARE_ANALYTICS_API_TOKEN # API token with Analytics:Read permission
CLOUDFLARE_ZONE_TAG # Zone ID for the PubPub traffic zone

If missing, the server logs a warning and the client shows a "not available" message instead of erroring.