Don't truncate color sampling mask to 8 bits

It's a 15 bit mask, so this was nonsense. Fixes the issues with smudge
colors going lighter or darker that required a fudge factor to be
applied.

extern/drawdance/libengine/dpengine/layer_content.c

@@ -416,12 +416,9 @@ static DP_UPixelFloat sample_dab_color(DP_LayerContent *lc, DP_BrushStamp stamp,
     alpha /= weight;
 
     // Unpremultiply, clamp against rounding error.
-    // The fudge factor avoids smudging causing the color to tend toward black.
-    // Not entirely sure why that happens, but fudging fixes it, so whatever.
-    float fudge = 0.005f;
-    red = CLAMP(red / alpha + fudge, 0.0f, 1.0f);
-    green = CLAMP(green / alpha + fudge, 0.0f, 1.0f);
-    blue = CLAMP(blue / alpha + fudge, 0.0f, 1.0f);
+    red = CLAMP(red / alpha, 0.0f, 1.0f);
+    green = CLAMP(green / alpha, 0.0f, 1.0f);
+    blue = CLAMP(blue / alpha, 0.0f, 1.0f);
 
     return (DP_UPixelFloat){
         .b = blue,

extern/drawdance/libengine/dpengine/paint.c

@@ -1059,8 +1059,9 @@ DP_BrushStamp DP_paint_color_sampling_stamp_make(uint16_t *data, int diameter,
             for (int x = 0; x < diameter; ++x) {
                 double dist = (DP_square_double(x - radius) + yy) * lut_scale;
                 int i = DP_double_to_int(dist);
-                *d = i < CLASSIC_LUT_SIZE ? DP_float_to_uint8(DP_BIT15 * lut[i])
-                                          : 0;
+                *d = i < CLASSIC_LUT_SIZE
+                       ? DP_float_to_uint16(DP_BIT15 * lut[i])
+                       : 0;
                 ++d;
             }
         }