Idea for testing: by default, read imports directly from notebook and populate requirements

[adrn]

For astropy-tutorials, we want to enable users to set up their environment to run a single notebook without installing packages/requirements needed for all notebooks in the repo. Right now, we have solved this by putting individual requirements.txt files along side each notebook tutorial. However, for most cases, we should be able to automatically figure out what the dependencies are for a given notebook by finding all import statements. We could then auto-generate a global requirements.txt file, and individual requirements.txt files for each notebook. Do you have any thoughts about this?