Refactor Perturbation Tests

[gabegma]

Separate the creation of the perturbation from the prediction part. This way, part of the cache will be reusable for different pipelines. This will also helps us provide something from the UI when there is no pipeline to export the perturbed dataset.

[Dref360]

Will probably get to it soon.

I can't assign myself, but we could add the self-assign action from HuggingFace https://github.com/huggingface/datasets/blob/master/.github/workflows/self-assign.yaml

[gabegma]

I added the self-assign GitHub action @Dref360!