Record UMA stats for Blink decoded image types

Source/platform/graphics/BitmapImage.cpp

Index: Source/platform/graphics/BitmapImage.cpp
diff --git a/Source/platform/graphics/BitmapImage.cpp b/Source/platform/graphics/BitmapImage.cpp
index e3a3a72d0a680ad01fc0836f3cdaa754899df775..d5fdd4973315845d1690bf16488ff22a6958dacf 100644
--- a/Source/platform/graphics/BitmapImage.cpp
+++ b/Source/platform/graphics/BitmapImage.cpp
@@ -340,6 +340,19 @@ bool BitmapImage::isSizeAvailable()
 
     m_sizeAvailable = m_source.isSizeAvailable();
 
+    if (m_sizeAvailable && (size().width() > 1 || size().height() > 1)) {

[Noel Gordon, 2015/08/01 02:33:06]

Yes better.  Let's define a static inline hasVisibleImageSize(size()) helper to
document the visible size thing.

In that helper, should it be the conjunction, viz., size().width() > N _&&_
size().height() > N ?

N=1 is obvious I suppose, but what value of N should we use to declare an image
visible?  N=6,8,10,...?  Consider the smallest box that could be visible _and_
meaningful to a user.

[urvang, 2015/08/04 20:21:16]

Done.
No, we require || operator because 1x100 as well as 100x1 sized images are
visible, for example.
Any image larger than 1x1 would be visible. So, N=1 is appropriate.

+        String fileExtention = m_source.filenameExtension();
+        DecodedImageType type =
+            fileExtention == "jpg"  ? ImageJPEG :
+            fileExtention == "png"  ? ImagePNG :
+            fileExtention == "gif"  ? ImageGIF :
+            fileExtention == "webp" ? ImageWebP :
+            fileExtention == "ico"  ? ImageICO :
+            fileExtention == "bmp"  ? ImageBMP :
+            ImageUnknown;
+        Platform::current()->histogramEnumeration("Blink.DecodedImageType", type, DecodedImageTypeMax + 1);
+    }
+
     return m_sizeAvailable;
 }