Add a benchmarking workflow to xclim

[coxipi]

Addressing a Problem?

xclim deals with big datasets and long computations. The difference between an optimal and suboptimal implementation of an algorithm can easily modify the running time by orders of magnitude. We should monitor more systematically the performance of our algorithms.

Potential Solution

We already use pytest, and this testing suite also offers benchmarking tools. It could be a good place to start.

https://pytest-benchmark.readthedocs.io/en/latest/

I will search if other big projects use this, and if so, how. Right now, I'm just thinking I could reproduce the "tests" folder with a "benchmarks" folder.

Additional context

No response

Contribution

• I would be willing/able to open a Pull Request to contribute this feature.

Code of Conduct

• I agree to follow this project's Code of Conduct

[Zeitsperre]

Relevant repository: https://github.com/Ouranosinc/xclim-benchmark

[SarahG-579462]

For notebook-level investigations into algorithms, I'd suggest taking a gander at flamegraph, https://github.com/H4dr1en/jupyterflame.git seems to work quite well