Test notebooks with a separate CI workflow

[mdpiper]

This PR moves the test-notebooks job to a separate workflow. I also disabled testing on pymt and terrainbento notebooks.

[mcflugen]

👍

[mcflugen]

This is the old way of reading an ESRI ASCII file. We should update this notebook to use the new way. But in another pull request.

[mdpiper]

Ah, OK. I couldn't remember which was the new way. I only changed it because it was failing in the CI test.

[mcflugen]

For a future pr, the new way is something like,

from landlab.io import esri_ascii

with open("data/square-test-basin.asc") as stream:
    grid = esri_ascii.load(filepath, name="topographic__elevation")
z = grid.at_node["topographic__elevation"]

[mcflugen]

Why did you decide to run this job as a separate workflow?

[mdpiper]

I like separating unit tests from checking notebooks. They feel like different things, even though we call both "tests" and they run through pytest. Also test.yml was longer than one screen.

Is there a penalty for splitting jobs across workflows instead of grouping them into one workflow?

[mdpiper]

Maybe this is a subject for discussion on the Forum!

[mcflugen]

Since the workflows are independent of one another, it's fine separating them. If you had a third job (e.g. maybe building/deploying docs or deploying to PyPI) that required both the regular tests and the notebook tests to complete, then you would have to put them in a single workflow.