perf: tune scalar argmin & argmax

[jvdd]

Continuation of #42

• tune scalar implementation of argmin and argmax

[codspeed-hq]

CodSpeed Performance Report

Merging #43 argminmax_opt (98673f6) will improve performances by 2.8%.

Summary

🔥 14 improvements
❌ 0 regressions
✅ 142 untouched benchmarks

🆕 0 new benchmarks
⁉️ 0 dropped benchmarks

Benchmarks breakdown

	Benchmark	main	argminmax_opt	Change
🔥	scalar_f32_argmax_rn	2.6 ms	2.2 ms	13.30%
🔥	scalar_u64_argmax	2.7 ms	1.5 ms	43.31%
🔥	scalar_u32_argmax	2.5 ms	1.3 ms	46.41%
🔥	scalar_f32_argmin_in	1.9 ms	1.7 ms	9.06%
🔥	scalar_f32_argmax_in	1.9 ms	1.7 ms	9.06%
🔥	scalar_i8_argmax	2.5 ms	1.3 ms	49.11%
🔥	scalar_u8_argmax	2.5 ms	1.3 ms	49.11%
🔥	scalar_i32_argmax	2.3 ms	1.2 ms	48.00%
🔥	scalar_u16_argmax	2.4 ms	1.3 ms	46.35%
🔥	scalar_f64_argmin_in	2.1 ms	1.9 ms	8.28%
🔥	scalar_f64_argmax_in	2.1 ms	1.9 ms	8.28%
🔥	scalar_i64_argmax	2.5 ms	1.4 ms	44.57%
🔥	scalar_f64_argmax_rn	2.7 ms	2.4 ms	12.44%
🔥	scalar_i16_argmax	2.4 ms	1.3 ms	46.36%