Tutorial text improvements

[alexreinking]

See the commit descriptions for the changes I made. A few highlights:

1. I audited all of the direct compiler invocations to make sure they work. All were misleading (e.g. you want -I to point to /path/to/include, not /path/to/Halide.h) and some were wrong (e.g. autoscheduler usage)
2. I removed the advice to use the Makefile, which contradicts our README.
3. I reworded some weirdly hostile language towards Apple's software.
4. I dropped OpenMP in favor of std::thread because OpenMP isn't readily available on macOS.
5. I changed the ancient SSE implementation in Lesson 9 to ARM Neon. I could have done AVX2, but it wasn't easy to test on my MBP (and Neon has the same register width anyway).

Breaking changes

None

Checklist

• Tests added or updated (not required for docs, CI config, or typo fixes)
• Documentation updated (if public API changed)
• Python bindings updated (if public API changed)
• Benchmarks are included here if the change is intended to affect performance.
• Commits include AI attribution where applicable (see Code of Conduct)

---

_{Stack created with GitHub Stacks CLI • Give Feedback 💬}

[alexreinking]

Ugh, nullglob strikes again.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 70.02%. Comparing base (e3e0ca0) to head (f7367c2).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #9348      +/-   ##
==========================================
- Coverage   70.11%   70.02%   -0.09%     
==========================================
  Files         259      259              
  Lines       79135    79135              
  Branches    19286    19286              
==========================================
- Hits        55487    55417      -70     
- Misses      17880    17899      +19     
- Partials     5768     5819      +51     

☔ View full report in Codecov by Harness.
📢 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[abadams]

Unfortunately the OpenMP replacement now reads like a bit of a strawman. Spinning up new threads just to run a single parallel loop is clearly going to be slow. It's not a convincing baseline.

[abadams]

Maybe we want std::for_each with a parallel execution policy?

[alexreinking]

Maybe we want std::for_each with a parallel execution policy?

This doesn't work because Apple libc++ doesn't have std::execution::par. Even if you enable the API with -fexperimental-library, the runtime serializes it anyway.

Unfortunately the OpenMP replacement now reads like a bit of a strawman. Spinning up new threads just to run a single parallel loop is clearly going to be slow. It's not a convincing baseline.

The halide_thread_pool.h implementation works as well as OpenMP on my machine, so I used that.