Add some end-to-end automated browser tests

I mentioned in this PR the steps I took to check manually a Django upgrade:

• Created a user
• Accessed a codelist for each of BNF, CTV3, dm+d, ICD-10, Read v2, SNOMED CT coding systems
  • Downloaded files from each of those codelists
  • Clicked around on everything for each codelist
• Uploaded a codelist for each of CTV3, ICD-10, Pseudo BNF, SNOMED CT coding systems
• Created and published a new codelist interactively
• Reset the user password
• Logged back in after password reset
• Clicked to the final index page
• Viewed the docs

This covers what I think are most of the features of OpenCodelists.

It may be worth having Playwright do this for us, so we can be more confident about upgrading dependencies without having to manually test.

Checking the behaviour of codelists, which is possibly more important, would require some way of providing sample codelist and coding system data.

Start with the manual workflow listed in the issue and assess Playwright as the entry point for automating it. Decide how sample codelist and coding-system data will be provided, then cover the listed user, codelist, password, documentation, and navigation flows; done means these checks run automatically.