feat: add Vertex AI and GCS backend support

[Xichun123]

Summary

• add a Vertex AI backend using Service Account authentication while preserving AI Studio support
• proxy Gemini model requests and GCS-backed Files API operations through the server
• add Docker profiles and a validated switch command for Vertex, server-managed AI Studio, and browser-managed BYOK modes
• inject authoritative runtime API routing so stale IndexedDB settings cannot keep the previous credential mode active after a profile switch
• document GCP IAM, GCS CORS/lifecycle setup, deployment variables, and backend switching in English and Chinese

Verification

• npm run typecheck
• npm run lint -- --no-cache
• npm test — 351 test files, 1909 tests passed
• npm run build
• npm run build:api
• shell syntax and all three Docker Compose profile configurations
• real Vertex AI countTokens request returned HTTP 200
• authenticated GCS metadata probe reached the configured bucket
• browser profile-switch validation for Vertex → BYOK → Vertex with stale IndexedDB settings
• npm audit --audit-level=high — currently blocked on main by its direct dompurify@^3.4.11 dependency conflicting with the existing dompurify@^3.4.9 override (EOVERRIDE); before rebasing onto that upstream change, the audit reported 0 vulnerabilities

Notes

• rebased onto current main (8d996841)
• no Service Account credentials, API keys, project IDs, or bucket names are committed
• AI Studio and BYOK profiles mount an empty secrets directory instead of the Vertex credential directory
• this PR intentionally excludes unrelated fork functionality and UI changes