Optimize Azure and Cosmos hosting costs

[sajeetharan]

Verified findings

• Cosmos uses manual shared throughput at 4,000 RU/s; average normalized utilization was 14.25% with 100% peaks
• impact-history contains about 956k documents and 1 GB with no TTL
• Container Apps averages 6.06% CPU and 29.25% memory with min replicas set to 1
• ACR Basic, Log Analytics 30-day retention, and Y1 Function plans are already cost-conscious

Changes

• set Container Apps min replicas to 0 while preserving max replicas at 3
• add a confirmation-gated, precondition-checked workflow for Cosmos autoscale max 4,000 RU/s and 120-day impact-history TTL
• reuse one CosmosClient across containers and migrate hot public routes to the shared helper
• cache country statistics and trending calculations for one hour
• patch the ip-address production advisory
• document rollout, tradeoffs, and seven-day follow-up checks

Safety

• this PR does not execute the one-time Cosmos workflow
• the workflow refuses unexpected throughput configurations and handles partial autoscale migration reruns
• search visibility remains uncached

Validation

• npm test (491 passed, 0 failed, 0 cancelled)
• npm run build
• npm audit --omit=dev (0 vulnerabilities)
• workflow YAML parse checks
• git diff --check