fix(tests): apply lowz cutoff to number density

[davecwright3]

Description

Fixes failing tests due to a low-z cutoff that wasn't applied uniformly.

Status

• Works on my machine
• Works on GitHub Actions

[davecwright3]

Actually, I think I can fix the notebook issues too. Let's not merge this just yet. I need to add a fix for a matplotlib deprecation in kalepy.

[davecwright3]

lzkelley/kalepy#21 is open. With this PR, the illustris notebook passes tests. I have another fix to relations.ipynb in 43c2e3e. After the kalepy PR is merged and there's a new release, I'll set the lower bound on kalepy to that release. Then, all tests, including the notebook tests, will be passing.

This PR has progressed beyond just fixing the redshift cutoff. I may delete it and create a new branch with an appropriate name. There is also still an issue to fix with the macOS-latest runners no longer coming with a bundled conda.

[davecwright3]

0fdd1a3 fixes the macOS issue.

[kayhangultekin]

@davecwright3 Is this ready now?

[davecwright3]

I'm still waiting on the kalepy bug fix. I'll need to set our kalepy version's lower bound after that. We could also just remove the kalepy plots from our test notebook and then the tests will be passing.

I think that's fine, kalepy is not holodeck so our tests really shouldn't depend on that codebase being up to date.

[kayhangultekin]

@davecwright3 which notebooks are causing the issue? Are the plots legit tests for which it would be good to have a working kde code to do the tests, or are they just helpful diagnostics?

[davecwright3]

It's the illustris_discrete notebook. The plots are just diagnostics.

[davecwright3]

I can replace the problem dist2d from kalepy with matplotlib, here's an example