fix(frontend): resolve production API URL fallback to localhost

Problem:
The unified Docker image builds the frontend at build time without
VITE_API_URL. Vite inlined import.meta.env.VITE_API_URL as undefined,
so every API call fell back to the hardcoded 'http://localhost:8080'.
This broke Casa deployments where the frontend loaded from the public
 domain but tried to reach the backend at localhost.

Solution:
1. Centralize API URL resolution in lib/api-url.ts via getApiOrigin().
   It checks runtime window.ENV first (injected by docker-entrypoint.sh
   at container startup), then build-time import.meta.env, then dev
   fallback. In production unified deployments it returns '' so API
   calls use same-origin relative URLs (/api/v1/...) that nginx proxies
   to the backend.
2. Replace all 50+ inline import.meta.env.VITE_API_URL || 'localhost'
   usages across 14 source files with getApiOrigin() / getApiV1BaseUrl().
3. Add build args and runtime sed substitution to Dockerfile and
   docker-entrypoint.sh so the same image works for any deployment.
4. Pass VITE_API_URL through docker-compose.yml and CI/CD build-args.

Verified:
- Production bundle contains 0 occurrences of localhost:8080.
- Container health check and /api/v1/auth/check-users both return 200.
- Runtime injection correctly sets VITE_API_URL='' for same-origin
  and VITE_API_URL='https://domain' for external backend.

Generated with [Devin](https://cli.devin.ai/docs)

Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>

Dockerfile

@@ -4,6 +4,14 @@
 # Stage 1: Build Frontend
 FROM node:22-alpine AS frontend-builder
 WORKDIR /app/frontend
+
+# Accept build arguments for Vite environment variables.
+# If unset, the frontend falls back to same-origin relative URLs in production.
+ARG VITE_API_URL
+ARG VITE_DEMO_MODE=false
+ENV VITE_API_URL=${VITE_API_URL}
+ENV VITE_DEMO_MODE=${VITE_DEMO_MODE}
+
 COPY frontend/package*.json ./
 RUN npm install
 COPY frontend/ ./