Add SSSE3 implementation for SkPMFloat, with faster get() and set().

src/core/SkPMFloat.h

 #ifndef SkPM_DEFINED
 #define SkPM_DEFINED
 
 #include "SkTypes.h"
 #include "SkColor.h"
 #include "Sk4x.h"
 
 // A pre-multiplied color storing each component as a float in the range [0, 255].
 class SK_STRUCT_ALIGN(16) SkPMFloat {
 public:
@@ skipping 27 matching lines @@
             && this->r() >= 0 && this->r() <= this->a()
             && this->g() >= 0 && this->g() <= this->a()
             && this->b() >= 0 && this->b() <= this->a();
     }
 
 private:
     // We mirror SkPMColor order only to make set()/get()/clamped() as fast as possible.
     float fColor[4];
 };
 
-#if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
+#if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSSE3
+    #include "../opts/SkPMFloat_SSSE3.h"
+#elif SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
     #include "../opts/SkPMFloat_SSE2.h"
 #elif defined(__ARM_NEON__)
     #include "../opts/SkPMFloat_neon.h"
 #else
     #include "../opts/SkPMFloat_none.h"
 #endif
 
 #endif//SkPM_DEFINED