[Tooling]: Added caching for Playwright (tests), and backend docker container when checking auto-generated API code

[rubenthoms]

Cache CI dependencies to speed up the frontend job

Summary

• Cache Playwright browser binaries across runs, keyed on frontend/package-lock.json. On a cache hit, only OS-level system dependencies are installed (install-deps); the ~300 MB browser download is skipped.
• Cache Docker build layers for the backend image used in the API sync check, using Docker Buildx's native GitHub Actions cache backend (type=gha,mode=max). Unchanged layers (e.g. the Poetry dependency install) are reused across runs; only layers whose inputs have changed are rebuilt.

Cache invalidation

Playwright — the cache key includes a hash of package-lock.json, so it is automatically busted whenever the Playwright version is bumped.

Docker — invalidation is content-addressed at the layer level by Docker itself. The COPY ./backend_py/primary instruction in the Dockerfile means any change to backend source files (routes, schemas, response models) will bust that layer and all subsequent ones, ensuring the API sync check always runs against the correct backend image.

Test plan

• Verify the Playwright cache is populated on the first run and restored on subsequent runs with no Playwright version change
• Verify the Docker layer cache is populated on the first run and partially restored (deps layer hit, source layer miss) after a backend source-only change
• Verify the API sync check correctly fails when the backend API changes but the generated frontend code has not been updated

[Copilot]

Pull request overview

This PR speeds up the frontend CI job by caching two large/expensive dependencies: Playwright browser binaries used by component tests, and Docker build layers for the backend image used during the “generated API code sync” check.

Changes:

• Add an actions/cache step to persist/restore Playwright browser binaries keyed by frontend/package-lock.json, and conditionally skip the browser download on cache hits.
• Switch the backend image build in the API sync check from docker build to docker buildx build with GitHub Actions cache backend (type=gha) to reuse Docker layers across runs.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[anders-kiaer]

Looks good - would be nice to improve frontend CI speed 👍

Some questions/comments below, otherwise LGTM.

[anders-kiaer]

I see we tried the buildx cache backend built for GitHub actions in an earlier commit (type=gha). What was the reason we wanted to change to local? Should we add a comment here stating the reason?

[anders-kiaer]

Is this needed? -e seems default on GitHub actions, and we do not use pipes in this job?

[rubenthoms]

You are right that -e is enabled by default and that we don't have any pipes. However, adding -u adds immediate value by adding fail-fast behaviour. -o pipefail is an insurance for future changes to the step as it discovers any hidden failures (e.g. some_command | grep ... where grep is successful but some_command fails). So yes, theoretically, we could only add -u but set -euo pipefail is a good default and immediately shows what is set (without having to know about and/or rely on GitHub).

[anders-kiaer]

Replacing this by something more explicit which also continues immediately when server is up?

Suggested change

	sleep 10
	# Waiting on OpenAPI server up and running:
	timeout 10 bash -c 'until curl -sf http://localhost:5000/openapi.json > /dev/null 2>&1; do sleep 1; done'

[anders-kiaer]

Good and more robust replacement of docker stop $CONTAINER_ID. We could also consider removing this completely as GitHub actions jobs are independent it should/will stop automatically when the job ends.