Migrate benchmarks

[HarryCampion]

This pull request reorganizes and modernizes the benchmarking and test utilities for the project. It moves benchmark analysis scripts and outputs to a unified location, updates Makefile targets to use uv run instead of direct Python calls, and eliminates the shared test_utils.py in favor of defining utilities directly in each test module. This improves maintainability, avoids import issues, and ensures all test scripts use up-to-date paths and conventions.

The most important changes include:

Benchmarking and Analysis Refactor:

• Moved analyze_benchmarks.py from tests/test_rhiza/benchmarks/ to .rhiza/utils/, updated it to use _benchmarks as the output directory, and adjusted all paths accordingly. [1] [2]
• Updated the Makefile (.rhiza/make.d/01-test.mk) so that benchmark outputs (JSON, HTML, histogram) are written to _benchmarks/ instead of inside the test directory. The analysis script is now run via uv run .rhiza/utils/analyze_benchmarks.py.
• Removed the obsolete tests/test_rhiza/benchmarks/README.md and .gitignore files, as well as the tests/conftest.py and test_utils.py utility modules, consolidating benchmark documentation and utilities. [1] [2] [3] [4]

Makefile and Test Target Modernization:

• Updated Makefile targets to use uv run pytest for running tests and generating coverage and reports, replacing direct calls to the Python interpreter in the virtual environment. Added checks for missing source or test folders, with clear warnings if they are absent. [1] [2] [3]
• Updated documentation in .rhiza/tests/README.md to reflect the new location for benchmarks (tests/benchmarks/).

Test Utilities Refactor:

• Removed the shared test_utils.py and instead defined helper functions (run_make, strip_ansi, setup_rhiza_git_repo) directly in each relevant conftest.py file for api and sync tests. This eliminates the need for relative imports and ensures all test modules are self-contained. [1] [2] [3] [4]
• Updated test imports and usage to reference the new utility locations, and adjusted test assertions to expect uv run pytest instead of .venv/bin/python -m pytest. [1] [2] [3] [4]

These changes collectively improve the structure, maintainability, and reliability of the project's test and benchmarking infrastructure.

[HarryCampion]

@tschm i will resolve conflicts but codefactor seems a bit buggy?

[tschm]

Yes, that's a weird CodeFactor problem. No issue and yet broken...