Make screenshots reproducible if possible

Why are we doing this?

It may be useful to rebuild the screenshots on a scheduled basis, or add a check to validate that screenshots are up-to-date in CI checks for branches.

Checking a binary diff is easy with git diff, but maybe the screenshots change, because of, for example:

• differences in Playwright behaviour across different operating systems
• differences in browser rendering behaviour following a Playwright update
• differences in test conditions (for example, dates/times — controllable, and maybe other aspects of codelist creation, such as version IDs)

Another comparison would be pixel comparison, rather than binary comparison, but again may also differ.

How will we know when it's done?

If we make the screenshot build reproducible, or we discover that this isn't possible, or it's not worth the effort.

What are we doing?

• Changing the way the screenshots work to remove variation between runs
• Investigating if the screenshots are reproducible
• Maybe trying to run taking screenshot in Docker, and maybe only via a CI workflow

Start by examining the existing screenshot build, Playwright test setup, and CI workflow. Compare screenshots across repeated runs and operating-system or Playwright changes, then evaluate Docker as a controlled environment. Done means screenshot generation is reproducible, or the investigation documents why it is not feasible or worthwhile.