Fix NLS mismatch in debug workspaces

[kylos101]

Description

Debug workspaces created via gp validate throw !!! NLS MISSING: 1837 !!! in the browser console, breaking UI components.

The IDE Docker image copies /vscode-web/ then /vscode-reh-linux-x64/ into /ide/. The REH overlay overwrites the web client's NLS files (nls.messages.js, etc.) with the REH server's version, which has different message counts and index ordering.

Normal workspaces are unaffected because blobserve reads only the first Docker layer (the web client). Debug workspaces serve from the merged /ide/ filesystem, so workbench.js references NLS indices that don't exist in the REH's nls.messages.js.

This became a problem after #21350 changed the build pipeline from compile-build-without-mangling (shared NLS state) to core-ci (separate builds with divergent NLS outputs).

Fix: restore the four web client NLS files after the REH overlay.

Related Issue(s)

Fixes CLC-2237

How to test

1. Build the IDE image with this change
2. Start a workspace and run gp validate
3. Open the debug workspace URL in the browser
4. Verify no NLS MISSING errors in the browser console

Documentation

No docs changes needed.

/hold

[kylos101]

Testing in a preview

[kylos101]

Fixed!

[kylos101]

I had to do the setup like so for the preview:

kubectl get configmap ide-config -o jsonpath='{.data.config\.json}' | \
  jq '.ideOptions.options.code.image = "eu.gcr.io/gitpod-dev-artifact/build/ide/code:kb-fix-debug-workspace-nls-mismatch-gha.249"' | \
  kubectl create configmap ide-config --from-file=config.json=/dev/stdin --dry-run=client -o yaml | \
  kubectl apply -f -

kubectl rollout restart deployment ide-service

And then start a new environment, and test gp validate.