chore: CI/CD improvements — frontend build check, health check, rollback

[saggacce]

Problemas que resuelve

1. Frontend no se reconstruía en producción — el build fallaba silenciosamente porque el CI usaba tsc --noEmit (root tsconfig) que no detectaba los mismos errores que tsc -b en apps/web
2. Sin health check post-deploy — un deploy roto se marcaba como ✅
3. Sin rollback — si algo fallaba, la app se quedaba en estado roto

Cambios en el pipeline

CI (pull_request + push a main):

• Añadido paso «Build frontend» que ejecuta npm run build en apps/web — misma compilación que producción
• Ahora los errores de build bloquean el merge antes de llegar al servidor

Deploy (push a main tras CI):

• Guarda el commit actual antes de actualizar
• Si el build falla → rollback al commit anterior inmediatamente
• Health check: 6 reintentos × 5s = 30s para que el servidor arranque
• Si el health check falla → rollback completo (git reset + rebuild + restart)
• El deploy solo marca ✅ si el servidor responde correctamente

i18n: Añadida clave nav.invitations en ES y EN — causaba el error de build silencioso

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: aa21df5aca

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Restore previous dist on frontend build rollback

When npm run build reaches the Vite phase and then fails, Vite prepares the output directory before writing (emptyOutDir defaults to true for an in-root dist), and apps/web/dist is gitignored while the running API serves that directory from disk in production. In that failure mode this rollback only resets tracked files, so the old PM2 process can continue serving an emptied or partially rewritten frontend until someone manually rebuilds the previous commit; rebuild the previous frontend (and restore deps if needed) before exiting this rollback path.

Useful? React with 👍 / 👎.