fix: prevent CDA GUI auth bootstrap crash

[krowvin]

Summary

• Load deployed Keycloak settings at runtime from the OpenAPI document instead of eagerly reading build-time Vite variables.
• Keep the CDA GUI available when OpenID configuration is missing or invalid, and show a controlled sign-in-unavailable state.
• Add a true application-level React error boundary and correct the React Router error fallback to use useRouteError().
• Share OpenID/Keycloak parsing between the application auth provider and Swagger UI, with focused regression tests.

Related Issue

Follow-up to #1734.

The user-list UI introduced an app-wide createKeycloakAuthMethod() call during main.jsx module evaluation. The deployable production-mode bundle does not define VITE_AUTH_HOST, so Groundwork received host: undefined and failed on host.trim() before React or the router error boundary could render. This left the CWBI dev UI on a white error page.

User Impact

CDA now derives authentication configuration from the deployment's own OpenAPI document. If the deployment does not advertise usable OpenID configuration, public CDA pages remain available and authenticated features explain that sign-in is unavailable instead of crashing the entire application.

Validation

• npm run test — 4 tests passed
• npm run lint
• npm run build using Vite production mode
• Repeated test, lint, and production build in an isolated npm ci --ignore-scripts installation using the locked Groundwork 3.9.0 dependency
• Loaded the production artifact locally in a browser with unavailable OpenID configuration; verified the CDA introduction page rendered, the sign-in-unavailable state appeared, the page was not blank, and the console had no errors

Checklist

• AI tools used

[krowvin]

I attempted to close and re-open this to retrigger the github actions that were stuck.

Ended up sorting an import in a file to retrigger the actions (committed)

[MikeNeilson]

I attempted to close and re-open this to retrigger the github actions that were stuck.

Ended up sorting an import in a file to retrigger the actions (committed)

In the future don't do that, we can manually trigger rerunning of failed actions.

[krowvin]

in future don't do that, we can manually trigger rerunning of failed actions.

I typically wouldn't but was hoping to get around the GitHub actions issues and thought i'd try it.

The manual retrigger was not working (due to the outage)