Incorrect `hpmean` correction in CIEDE2000 implementation

[michel-leonard]

Hi,
While validating the DeltaE function of this package against several established implementations (including colormath (Python), palette (Rust), and my own implementation at michel-leonard/ciede2000-color-matching), I noticed a systematic deviation in go-colorful’s results.

Here, the root cause is in the handling of the mean hue angle (hpmean) when the absolute difference between hp1 and hp2 exceeds 180°:

if math.Abs(hp1-hp2) > 180 {
	if hp1+hp2 < 360 {
		hpmean += 180
	} else {
		hpmean -= 180
	}
}

This condition incorrectly biases the angular correction and leads to a noticeable precision error, especially in edge cases near the 0°/360° boundary.

Fix:
According to the original paper by Sharma et al. (2005), the correct approach is simply:

if math.Abs(hp1-hp2) > 180 {
	hpmean += 180
}

This change alone brings the precision of the DeltaE function back in line with the reference implementations (error < 1e-12). I've tested this over 100 million randomized samples with perfect numerical consistency.

Please let me know if you'd like me to submit a PR with the fix and a corresponding unit test.

Thanks for maintaining this useful library!

Best,
Michel

[makew0rld]

Thanks for looking into this! A PR would definitely be appreciated, and we can discuss more there if needed.