Flaky tests on GitHub actions

[KasperFyhn]

There seems to be some inconsistency in whether the tests in GitHub actions pass or not due to disk space issues. Sometimes they fail, sometimes they do not.

It possibly has to do with size of models being downloaded, but it is not 100% sure. No matter what, there are some circumstances that can lead to failing tests randomly. At the time of writing, I am suspecting race conditions, concurrency issues or the like.