Run examples and Jupyter notebooks in CI

[EwoutH]

It would be useful to run all the Python and Jupyter notebook examples in CI periodically, to see if they still work with the latest versions of packages, and as integration testing for ourselves.

Dumping this workflow here that I just wrote for another repo, will port it to the EMAworkbench later this week.

name: Execute Jupyter Notebooks

on:
  push:
    paths: ['Code/**.ipynb', '**/run_notebooks.yml']
  pull_request:
    paths: ['Code/**.ipynb', '**/run_notebooks.yml']
  workflow_dispatch:
  schedule:
    - cron: '0 6 1 * *'  # Runs at 06:00 UTC 1st day of the month

jobs:
  run-notebooks:
    runs-on: ubuntu-latest
    steps:
    - uses: actions/checkout@v4
    - uses: actions/setup-python@v5
    - name: Install Dependencies
      run: |
        pip install nbconvert ipykernel
        pip install numpy matplotlib biogeme pandas seaborn scikit-learn
    - name: Find and Execute Jupyter Notebooks
      run: |
        find Code -name "*.ipynb" -print0 | while IFS= read -r -d $'\0' notebook; do
          # Save the current directory
          current_dir=$(pwd)
          # Extract the directory path of the notebook
          notebook_dir=$(dirname "$notebook")
          # Change to the notebook directory
          cd "$notebook_dir"
          # Execute the notebook
          if jupyter nbconvert --to notebook --execute "$(basename "$notebook")" --output "$(basename "$notebook")"; then
            echo -e "\u2705 Passed: $notebook"
          else
            echo -e "\u274C Failed: $notebook"
          fi
          # Change back to the original directory
          cd "$current_dir"
        done

[quaquel]

This is becoming a bit more important with the overhaul I am doing.

At a minimum, for integration testing, it would be really helpful if a small subset of examples is run in CI. At a minimum, we need to rung a single perform_experiments and optimize example.

The notebooks are also relevant, but on the documentation side of things.

[EwoutH]

A few options:

1. jupyter nbconvert --execute to see if it just runs
2. pytest-nbval for regression testing. This runs notebooks and compares outputs against saved versions:

   - run: pip install pytest nbval
   - run: pytest --nbval-lax Code/

3. nbmake for running notebooks with pytest integration.

   - run: pip install nbmake
   - run: pytest --nbmake Code/

What do you think is most useful here?

[quaquel]

I'll take a look at these options at some later point so no clear idea yet.

My hunch is that the main thing is to make sure they just run, but some regression testing might be nice as well. However, that requires that the seed control is in place fully first.

I have added 3 examples for integration testing in #413