feat: mix colors in the linear sRGB model

This seems to produce results that are closer to perceptual models.
See https://bottosson.github.io/posts/colorwrong/

packages/keybr-color/lib/mix.test.ts

@@ -8,7 +8,7 @@ test("mix", () => {
   const b = hexColor(0xff8000);
   equal(mixColors(a, b, -100.0).hex(), 0x0080ff);
   equal(mixColors(a, b, 0.0).hex(), 0x0080ff);
-  equal(mixColors(a, b, 0.5).hex(), 0x808080);
+  equal(mixColors(a, b, 0.5).hex(), 0xbc80bc);
   equal(mixColors(a, b, 1.0).hex(), 0xff8000);
   equal(mixColors(a, b, 100.0).hex(), 0xff8000);
 });

packages/keybr-color/lib/mix.ts

@@ -1,16 +1,23 @@
 import { clamp } from "@keybr/lang";
 import { RgbColor } from "./color-rgb.ts";
+import { rgbGammaToLinear, rgbLinearToGamma } from "./convert-xyz.ts";
 import { type Rgb } from "./types.ts";
 
+const u: Rgb = { r: 0, g: 0, b: 0, alpha: 1 };
+const v: Rgb = { r: 0, g: 0, b: 0, alpha: 1 };
+
 export function mixColors(
   a: Readonly<Rgb>,
   b: Readonly<Rgb>,
-  n: number,
+  r: number,
 ): RgbColor {
-  n = clamp(n, 0, 1);
-  return new RgbColor(
-    (b.r - a.r) * n + a.r,
-    (b.g - a.g) * n + a.g,
-    (b.b - a.b) * n + a.b,
-  );
+  r = clamp(r, 0, 1);
+  rgbGammaToLinear(a, u);
+  rgbGammaToLinear(b, v);
+  u.r = (v.r - u.r) * r + u.r;
+  u.g = (v.g - u.g) * r + u.g;
+  u.b = (v.b - u.b) * r + u.b;
+  u.alpha = (v.alpha - u.alpha) * r + u.alpha;
+  rgbLinearToGamma(u, v);
+  return new RgbColor(v.r, v.g, v.b, v.alpha);
 }