Add CanvasColorType, with unorm8 and float16 support

The default behavior is unorm8, which matches the existing behavior of
all user agents.

The name unorm8 is chosen to match modern APIs which distinguish between
8-bit unsigned normalized and 8-but unsigned integer (e.g,
VK_FORMAT_R8G8B8A8_UNORM vs VK_FORMAT_R8G8B8A8_UINT in Vulkan,
MTLPixelFormatRGBA8Unorm vs MTLPixelFormatRGBA8Unorm in Metal, and
"rgba8unorm" vs "rgba8uint" in WebGPU).

The use of colorType (instead of pixelFormat) is used to avoid the
complexity of selecting channel ordering (BGRA vs RGBA) and interactions
with the alpha parameter (RGB, RGBX, or BGRX potentially being required if
alpha is false).

source

@@ -65366,12 +65366,15 @@ typedef (<span>HTMLOrSVGImageElement</span> or
 
 enum <dfn enum>PredefinedColorSpace</dfn> { "<span data-x="dom-PredefinedColorSpace-srgb">srgb</span>", "<span data-x="dom-PredefinedColorSpace-display-p3">display-p3</span>" };
 
+enum <dfn enum>CanvasColorType</dfn> { "<span data-x="dom-CanvasColorType-unorm8">unorm8</span>", "<span data-x="dom-CanvasColorType-float16">float16</span>" };
+
 enum <dfn enum>CanvasFillRule</dfn> { "<span data-x="dom-context-2d-fillRule-nonzero">nonzero</span>", "<span data-x="dom-context-2d-fillRule-evenodd">evenodd</span>" };
 
 dictionary <dfn dictionary>CanvasRenderingContext2DSettings</dfn> {
   boolean <span data-x="dom-CanvasRenderingContext2DSettings-alpha">alpha</span> = true;
   boolean <span data-x="dom-CanvasRenderingContext2DSettings-desynchronized">desynchronized</span> = false;
   <span>PredefinedColorSpace</span> <span data-x="dom-CanvasRenderingContext2DSettings-colorSpace">colorSpace</span> = "<span data-x="dom-PredefinedColorSpace-srgb">srgb</span>";
+  <span>CanvasColorType</span> <span data-x="dom-CanvasRenderingContext2DSettings-colorType">colorType</span> = "<span data-x="dom-CanvasColorType-unorm8">unorm8</span>";
   boolean <span data-x="dom-CanvasRenderingContext2DSettings-willReadFrequently">willReadFrequently</span> = false;
 };
 
@@ -65820,6 +65823,19 @@ context.fillRect(100,0,50,50); // only this square remains</code></pre>
 
   <hr>
 
+  <p>The <code>CanvasColorType</code> enumeration is used to specify the <span
+  data-x="concept-canvas-color-type">color type</span> of the canvas's backing store.</p>
+
+  <p>The "<dfn enum-value for="CanvasColorType"><code
+  data-x="dom-CanvasColorType-unorm8">unorm8</code></dfn>" value indicates that the type
+  for all color channels is 8-bit unsigned normalized.</p>
+
+  <p>The "<dfn enum-value for="CanvasColorType"><code
+  data-x="dom-CanvasColorType-float16">float16</code></dfn>" value indicates that the type
+  for all color channels is 16-bit floating point.</p>
+
+  <hr>
+
   <p>The <code>CanvasFillRule</code> enumeration is used to select the <dfn>fill rule</dfn>
   algorithm by which to determine if a point is inside or outside a path.</p>
 
@@ -65962,6 +65978,12 @@ context.fillRect(100,0,50,50); // only this square remains</code></pre>
   data-x="concept-canvas-color-space">color space</span> indicates the color space for the
   <span>output bitmap</span>.</p>
 
+  <p>The <code>CanvasSettings</code> object also has a <dfn
+  data-x="concept-canvas-color-type">color type</dfn> setting of type
+  <code>CanvasColorType</code>. The <code>CanvasSettings</code> object's <span
+  data-x="concept-canvas-color-type">color type</span> indicates the data type of the
+  color and alpha components of the pixels of the <span>output bitmap</span>.</p>
+
   <p>To <dfn data-x="canvas-setting-init-bitmap">initialize a <code>CanvasSettings</code> output
   bitmap</dfn>, given a <code>CanvasSettings</code> <var>context</var> and a
   <code>CanvasRenderingContext2DSettings</code> <var>settings</var>:</p>
@@ -65980,6 +66002,10 @@ context.fillRect(100,0,50,50); // only this square remains</code></pre>
    <var>settings</var>["<dfn dict-member for="CanvasRenderingContext2DSettings"><code
    data-x="dom-CanvasRenderingContext2DSettings-colorSpace">colorSpace</code></dfn>"].</p></li>
 
+   <li><p>Set <var>context</var>'s <span data-x="concept-canvas-color-type">color type</span> to
+   <var>settings</var>["<dfn dict-member for="CanvasRenderingContext2DSettings"><code
+   data-x="dom-CanvasRenderingContext2DSettings-colorType">colorType</code></dfn>"].</p></li>
+
    <li><p>Set <var>context</var>'s <span data-x="concept-canvas-will-read-frequently">will read
    frequently</span> to <var>settings</var>["<dfn dict-member
    for="CanvasRenderingContext2DSettings"><code
@@ -71832,6 +71858,16 @@ interface <dfn interface>OffscreenCanvasRenderingContext2D</dfn> {
   data-x="dom-canvas-toBlob">toBlob()</code> method to the canvas, given the appropriate dimensions,
   has no visible effect beyond, at most, clipping colors of the canvas to a more narrow gamut.</p>
 
+  <p>For image types that support multiple bit depths, the serialized image must use the bit depth
+  that best preserves content of the underlying bitmap.</p>
+
+  <p class="example">For example, when serializing a 2D context that has
+  <span data-x="concept-canvas-color-type">color type</span> of
+  <span data-x="dom-CanvasColorType-float16">float16</span> to <var>type</var>
+  "<code>image/png</code>", the resulting image would have 16 bits per sample.
+  This serialization will still lose significant detail (all values less than 0.5/65535
+  would be clamped to 0, and all values greater than 1 would be clamped to 1).</p>
+
   <p>If <var>type</var> is an image format that supports variable quality (such as
   "<code>image/jpeg</code>"), <var>quality</var> is given, and <var>type</var> is not
   "<code>image/png</code>", then, if <var>quality</var> <span data-x="js-Number">is a Number</span>