fix(met-web): Investigate Openshift/github action deployments

[tom0827]

Addresses the intermittent whitescreen seen on admin sign-in after dev/test deploys.

• nginx: serve /assets/ from its own location with try_files $uri =404. Previously a missing content-hashed chunk fell through to index.html and returned 200 with Content-Type: text/html, so the browser parsed HTML as a module script and failed deep in an unrelated vendor chunk.
• nginx: Cache-Control no-cache on index.html, the asset manifest, so a client cannot act on a manifest older than the running pod.
• regenerate package-lock.json; it still pinned met-formio 2.0.0-rc1 while package.json asked for ^2.0.3.
• Dockerfile and CI: npm install -> npm ci, so lock drift fails the build instead of being silently re-resolved. Also makes builds reproducible; two clean builds now emit identical asset filenames.