tsdownsample: extremely fast time series downsampling written in Rust

tsdownsample brings highly optimized time series downsampling to Python, by using the SIMD optimized argminmax crate \- which matches or even outperforms numpy's speed (300x faster for f16). Through using runtime feature set detection, this crate selects the most optimal SIMD implementation for the current CPU (supports SSE, AVX(2), AVX512, NEON).

As a result, tsdownsample is quite fast! By using Rayon to parallellize the downsampling, tsdownsample can downsample 500 MILLION datapoints (f32) in 0.05s!

Code: https://github.com/predict-idlab/tsdownsample

I would love to hear your feedback on this!