perf(delta): O(n) cumulative sum instead of O(n w) per-element np.mean

[cbueth]

This PR replaces the O(n w) per-element np.mean loop in delta detrending with a single np.cumsum pass. Local tests show it is 37 to40 times faster (38µs -> 1µs). I added more tests with hardcoded expected values, a naive reference implementation, and constant-series symmetry tests. Through this we can assure the functionality is not breaking.

[codspeed-hq]

Merging this PR will improve performance by ×2.5

⚡ 4 improved benchmarks
✅ 24 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	test_detrend_method[delta-w100]	17.6 ms	3.5 ms	×5.1
⚡	test_detrend_method[delta-w10]	17.4 ms	3.5 ms	×5
⚡	test_detrend_method[z_score-p10]	2.1 ms	1.7 ms	+22.45%
⚡	test_ordinal_patterns	15.9 ms	13.4 ms	+18.17%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing perf/delta-convolution (6af1d81) with main (875adcd)}

[codecov]

Welcome to Codecov 🎉

Once you merge this PR into your default branch, you're all set! Codecov will compare coverage reports and display results in all future pull requests.

Thanks for integrating Codecov - We've got you covered ☂️