Fix override.aes size to linewidth for linetype legend (#449)

[ishaan-arora-1]

Fixes the override.aes in the linetype legend for mcmc_trace() divergence plots — changes size to linewidth, which is the correct aesthetic for line widths since ggplot2 3.4.0.

The stat_function -> geom_function change has been reverted per review feedback.

[jgabry]

Thanks. this seems like a change we should make but we have a failing test:

https://github.com/stan-dev/bayesplot/actions/runs/22910316067/job/66500925400?pr=450#step:7:1175

[ishaan-arora-1]

@jgabry only ubuntu fails, we can merge now right?

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.55%. Comparing base (b1cb851) to head (64bd86c).
⚠️ Report is 10 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #450   +/-   ##
=======================================
  Coverage   98.55%   98.55%           
=======================================
  Files          35       35           
  Lines        5864     5864           
=======================================
  Hits         5779     5779           
  Misses         85       85           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jgabry]

stat_function(), which has been superseded by geom_function() since ggplot2 3.3.0

Actually, I have a question about this. I just checked the doc for stat_function and it's not deprecated and it doesn't say that geom_function should be used instead. Where do you see that it has been superseded? Maybe I didn't look in the right place.

[ishaan-arora-1]

overlay_function() was wrapping stat_function(), which has been superseded by geom_function() since ggplot2 3.3.0. Switched it over —

Hey @jgabry, to reframe the pr, instead of 'superseded' i should say 'recommended alternative' .
the ggplot2 3.3.2 states : 'A newly added geom_function() is now recommended to use in conjunction with/instead of stat_function()'

you should find it here exactly : https://ggplot2.tidyverse.org/news/index.html#ggplot2-332

[jgabry]

the API is compatible since we just pass ... through

I don't think this is fully true. For example stat_function(..., geom = "point", inherit.aes = FALSE) produces a point layer whereas geom_function(..., geom = "point", inherit.aes = FALSE) warns that geom is unknown and keeps the default function geom. Since overlay_function() exposes ... this would technically be breaking backwards compatibility. I'm not sure if anyone is actually trying to specify geom, so maybe not a big deal, but I think since stat_function isn't actually deprecated we should avoid making this change.

We don't need to close this PR though, we can keep the size -> linewidth fix if you remove the stat_function change.

[jgabry]

Looks good thanks!