Run tutorials on latest Haystack version + on main

[bilgeyucel]

Fixes #197

Updates

• Add generate_matrix.py to generate the tutorial matrix
  • This script creates a list item for each tutorial (latest Haystack version + main) e.g.

     [{"notebook": "01_Basic_QA_Pipeline", "haystack_version": "v1.18.1"}, 
     {"notebook": "01_Basic_QA_Pipeline", "haystack_version": "main"},...]

  • Later on, we use the haystack_version info to pick the correct docker image to run the test
  • Skip some tutorials when they don't have a runnable Python code (tutorial 20) or need GPU to run (e.g. tutorial 2)
• Update nightly.yml to get the matrix from the generate_matrix.py script
• Mark some tutorials as needs_gpu and skip the nightly tests for these tutorials
• Change the way tutorial 25 handles the API Key

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[silvanocerza]

I ran this version of the workflow and all the tutorials that were expected to run did. Some failed but that's expected, we'll fix them in a later PR. Workflow run: https://github.com/deepset-ai/haystack-tutorials/actions/runs/5643922044

It took a lot of time since we're running only 2 jobs at a time to avoid draining runners for the whole org. I don't think it's an issue since these are nightly and there's no haste for those.

All in all the workflow logic is sound and works as expected. We'll expand on that to handle tutorial extra dependencies in a separate PR.