Benchmark between different documentation build options - #2562

This PR is a benchmark for three different setups for fresh documentation builds for PlasmaPy on GitHub Actions with ubuntu-latest:

• Build docs, regular, serial: 172 s (baseline)
• Build docs, regular, parallel (with -j auto): 187 s (109% of baseline)
• Build docs, skip notebooks, serial: 143 s (83% of baseline)

Running sphinx-build in parallel makes it take 9% longer than running it serially!

My conclusions are:

• We shouldn't parallelize PlasmaPy's doc builds — at least for now.
• My main hypotheses for the reasons why doc builds are slightly slower in parallel are:
  1. A limitation of sphinx-build
  2. A Sphinx extension or something about PlasmaPy's config is blocking possible performance gains from parallel doc builds.
• Skipping example notebook builds will save ∼30 s. We should consider if that is worth it, keeping in mind that the notebook builds will still be done by Read the Docs.
• Astral needs to set up a bat-signal so we can get them to rewrite Sphinx in Rust.
• Bat signals probably don't work that well on clear nights or during the day or in space and they've probably been obsoleted by cell phones anyway.