Implement a speedup for Y-only blurs by transposing.

src/opts/opts_check_SSE2.cpp

Index: src/opts/opts_check_SSE2.cpp
diff --git a/src/opts/opts_check_SSE2.cpp b/src/opts/opts_check_SSE2.cpp
index efbdbc1850894bdcc7be170fd0a54a62b9e290e6..aaf6b2ef824dfab2e7b26f261862b7f97614a145 100644
--- a/src/opts/opts_check_SSE2.cpp
+++ b/src/opts/opts_check_SSE2.cpp
@@ -271,14 +271,15 @@ SkMorphologyProc SkMorphologyGetPlatformProc(SkMorphologyProcType type) {
 
 bool SkBoxBlurGetPlatformProcs(SkBoxBlurProc* boxBlurX,
                                SkBoxBlurProc* boxBlurY,
-                               SkBoxBlurProc* boxBlurXY) {
+                               SkBoxBlurProc* boxBlurXY,
+                               SkBoxBlurProc* boxBlurYX) {
 #ifdef SK_DISABLE_BLUR_DIVISION_OPTIMIZATION
     return false;
 #else
     if (!cachedHasSSE2()) {
         return false;
     }
-    return SkBoxBlurGetPlatformProcs_SSE2(boxBlurX, boxBlurY, boxBlurXY);
+    return SkBoxBlurGetPlatformProcs_SSE2(boxBlurX, boxBlurY, boxBlurXY, boxBlurYX);
 #endif
 }