feat: add Platform-managed AI and storage starter

[xiongxz]

Summary

• publish the existing chatbot template as AI Chat + Storage Starter with ai.chat, ai.streaming, and storage capability requirements
• route chat exclusively through InsForge and omit model by default so the runtime-owned AI_DEFAULT_MODEL selects DeepSeek, LiteLLM, or another configured OpenAI-compatible model
• keep uploads behind the InsForge Storage SDK so applications never receive COS credentials
• remove Vercel AI Gateway/BYOK configuration and limit generated app configuration to the public InsForge endpoint and anon key
• validate capability declarations in the registry, add coverage, and prevent the Lexmount fork from calling InsForge's production marketplace sync workflow
• upgrade Next.js and transitive dependencies to versions with a clean production audit

Validation

• cd scripts && npm test
• cd scripts && node validate-registry.mjs ../registry.json ..
• cd chatbot && npm run typecheck
• cd chatbot && NEXT_PUBLIC_INSFORGE_URL=https://example.invalid NEXT_PUBLIC_INSFORGE_ANON_KEY=test-anon npm run build
• cd chatbot && npm audit --omit=dev (0 vulnerabilities)
• go test ./internal/templatecatalog ./internal/platform in lexmount/insforge-platform
• git diff --check

After merge

Configure Insight Flow's OFFICIAL_TEMPLATE_REPOSITORY as https://github.com/lexmount/insforge-templates, run an official template sync, review/publish the imported template, then create one v2 application for live AI streaming and COS upload/download/isolation acceptance.