🙈 change default minmax_ratio to 4

[jvdd]

Changes the minmax_ratio to 4 (was previously 30) - this change is based on the empirical evidence gathered in our research on the MinMaxLTTB work; https://arxiv.org/abs/2305.00332 (in this paper minmax_ratio is called the "preselection ratio").

Note that in plotly-resasampler we already used 4 as default value (in the wrapper for this package);
https://github.com/predict-idlab/plotly-resampler/blob/741da5d2358bee59c59d1591b0722ab0ea66bbb6/plotly_resampler/aggregation/aggregators.py#L211

This PR also fixes a minor bug when calling MinMaxLTTB.downsample (without an x array) -> the indices of the preselected data points should be passed to the LTTB algorithm (2nd step of the algorithm) in order to have a more correct triangle surface calculation (-> will resemble more closesly vanilla LTTB behavior).

---

Credits are due to @raphaelvallat for uncovering this issue

[codspeed-hq]

CodSpeed Performance Report

Merging #62 will degrade performances by 92.12%

_{Comparing minmax_ratio (aa937c5) with main (25d5648)}

Summary

⚡ 174 improvements
❌ 23 regressions
✅ 193 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	main	minmax_ratio	Change
⚡	test_minmax_no_x[False-float32-100-100,000]	8.1 ms	1.6 ms	×5.1
⚡	test_minmax_no_x[False-float32-100-1,000,000]	69.4 ms	4.6 ms	×15
⚡	test_minmax_no_x[False-float32-5,000-100,000]	9.3 ms	2.2 ms	×4.2
⚡	test_minmax_no_x[False-float32-5,000-1,000,000]	71.5 ms	6.1 ms	×12
⚡	test_minmax_no_x[False-float64-100-1,000,000]	139.8 ms	9 ms	×16
⚡	test_minmax_no_x[False-float64-100-100,000]	15 ms	1.9 ms	×7.8
⚡	test_minmax_no_x[False-float64-5,000-100,000]	16.2 ms	2.6 ms	×6.2
⚡	test_minmax_no_x[False-float64-1,000-100,000]	14.8 ms	1.6 ms	×9.3
⚡	test_minmax_no_x[False-float64-5,000-1,000,000]	141.6 ms	10.2 ms	×14
⚡	test_minmax_no_x[True-float32-100-1,000,000]	69.6 ms	5 ms	×14
⚡	test_minmax_no_x[True-float32-100-100,000]	7.6 ms	1.3 ms	×6
⚡	test_minmax_no_x[True-float32-1,000-1,000,000]	70.1 ms	5.6 ms	×13
⚡	test_minmax_no_x[False-float32-1,000-1,000,000]	69.8 ms	4.9 ms	×14
⚡	test_minmax_no_x[True-float32-5,000-100,000]	9.8 ms	2.9 ms	×3.4
⚡	test_minmax_no_x[True-float64-100-100,000]	14.7 ms	1.8 ms	×8.1
⚡	test_minmax_no_x[True-float64-100-1,000,000]	140.1 ms	9.6 ms	×15
⚡	test_minmax_no_x[False-float32-1,000-100,000]	7.7 ms	1.1 ms	×7.1
⚡	test_minmax_no_x[True-float64-5,000-1,000,000]	142.1 ms	10.8 ms	×13
❌	test_minmax_no_x[True-int32-1,000-100,000]	2 ms	2.3 ms	-14.02%
⚡	test_minmax_no_x[True-float64-5,000-100,000]	16.6 ms	3.3 ms	×5
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.