CI runtime

[Meier-Matthias]

As the runtime is capped at 6 hours, we should test the chapters one by one. Maybe we run the tests on a local server (and use the output as bookdown).

[MaximilianPi]

Is there a way to split the CI with github actions? I suggest to remove the large models from the CI for now (e.g. the transfer learning, augmentation, AE, VAE, and the GANs). To be honest, I included the transfer learning only in the CI to see if it is feasible.
Testing locally should be the last resort (we don't have much local resources)

[Meier-Matthias]

Well, I don't think that there is a convenient way to split. Just take the tests in and out manually.
Doing continuous testing locally is a waste of resources. But the last test should be done this way (and the outputs used for the bookdown), I think.