A/B testing for notebook results

[SimonHeybrock]

• This would be useful outside scipp.
• Create, e.g., 2 conda envs, run notebooks, store tagged results to files
  • Input data would typically be obtained by the notebook using pooch. Could inject other files somehow, e.g., to test with a number of different inputs.
• Load files, compare (with tolerance)
• Would be most useful if adding a new "test" would be as simple as adding a line (filename) to a list
• When and where should this run?

[nvaytet]

• Add two owners per workflow? one developer for the code itself, one scientist who has the knowledge to approve the scientific results