Add CodSpeed performance benchmarks

[codspeed]

This PR sets up continuous performance tracking with CodSpeed using pytest-codspeed.

What's included

Benchmarks (benchmarks/, 15 in total, run against a real bpy in CPU simulation mode):

• test_bench_builder.py: building trees with the builder API. Covers the README example (with and without the Sugiyama arrange step), a 4-bit boolean decoder made of many BooleanMath nodes, a long chain of math operator expressions, and a small shader tree.
• test_bench_layout.py: nodebpy.arrange on pre-built trees, using both the sugiyama and simple methods. The tree is built outside the measured region, so only the layout work is timed.
• test_bench_codegen.py: to_python() export (with format=False so ruff isn't part of the measurement) for each sample tree, plus running the generated code to rebuild the tree.

The benchmarks have their own conftest.py. It reloads tests/data/test_startup.blend before each benchmark and skips the save-to-.blend teardown that the test suite does.

CI: .github/workflows/codspeed.yml runs uv run pytest benchmarks/ --codspeed through CodSpeedHQ/action@v5 in simulation mode on ubuntu-latest. It triggers on pushes to main, on pull requests, and on workflow_dispatch, and authenticates with OIDC, so no token secret is needed.

Other changes: added pytest-codspeed to the dev dependency group (uv.lock is updated) and added a CodSpeed badge to the README.

Running locally

uv sync --all-extras
uv run pytest benchmarks/            # quick run as ordinary tests
uv run pytest benchmarks/ --codspeed # benchmark mode

Next steps

• Once this is merged, CodSpeed will record a baseline on main and post performance reports on later pull requests.
• The Sugiyama layout benchmarks are the slowest ones and the most useful for spotting layout regressions. More trees from tests/layout/corpus can be added if you want broader layout coverage.

[codspeed]

Congrats! CodSpeed is installed 🎉

🆕 15 new benchmarks were detected.

You will start to see performance impacts in the reports once the benchmarks are run from your default branch.

Detected benchmarks

• test_build_decoder_4bit: 204.3 ms
• test_build_math_expression: 501.2 ms
• test_build_readme_tree_no_arrange: 23 ms
• test_build_readme_tree_sugiyama: 64.8 ms
• test_build_shader_tree: 18.5 ms
• test_to_python_roundtrip: 68.9 ms
• test_to_python[decoder_4bit]: 47.5 ms
• test_to_python[math_expression]: 80.5 ms
• test_to_python[readme]: 14.6 ms
• test_to_python[shader]: 16.4 ms
• test_arrange_decoder_4bit[simple]: 1.7 s
• test_arrange_decoder_4bit[sugiyama]: 5.3 s
• test_arrange_math_expression: 2.4 s
• test_arrange_readme_tree[simple]: 35.7 ms
• test_arrange_readme_tree[sugiyama]: 42.1 ms

---

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.30%. Comparing base (107ed3c) to head (8e567d4).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #253   +/-   ##
=======================================
  Coverage   98.30%   98.30%           
=======================================
  Files         101      101           
  Lines       30574    30574           
=======================================
  Hits        30056    30056           
  Misses        518      518           

☔ 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.