This is just a small refactoring to make it easier to adjust this method.

… because that is what it actually does -- it doesn’t load a page from the document, it loads a list of pages that contain the search terms.

Fixes alephdata#3320

I’m not sure what exactly the intention here was originally, but it didn’t have any effect in its current version. `page` and `res.index` would always be undefined and as a result, the `active` class was applied to every single link. The `active` class doesn’t apply any styles.

Probably this is just leftover from a previous iteration of the UI, so I’ve removed this.

Fixes alephdata#3999.

This is related to alephdata#3948 in which I fixed the infinite request loop in case requests to the metadata API fail. This works as intended, but has one unintended side effect: We were actually kind of relying on the previous behavior to handle expired session tokens.

For context: When a user logs in, the session token is stored. When the session token expires, future API requests using that session token will obviously fail. The UI was previously handling 401 response codes to invalidate the stored session token and display a login popup or redirect to the OAuth service.

The change introduced in alephdata#3948 also prevented the handling of requests that failed to expired session tokens. I’m not convinced that this way of handling expired session tokens is a good solution as it’s pretty opaque.

So with this change, we will keep the newly introduced behavior (which shows an error message and a "Retry" button) in case a request to the metadata API fails, *except* if it failed with a 401 response in which case we still do whatever happened before.

`postgres:latest` currently brings in postgres 18, which has an intentional breaking change regarding its `PGDATA` folder: https://hub.docker.com/_/postgres#pgdata. It seems sensible to mount `/var/lib/postresql` in this guide.

chore: Update example docker-compose for pg 18

chore: typo & casing for consistency