Add Load2/store2 to Sk4x (dumb impl for now)

src/core/Sk4x.h

 #ifndef Sk4x_DEFINED
 #define Sk4x_DEFINED
 
 #include "SkTypes.h"
 
 #define SK4X_PREAMBLE 1
     #if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
         #include "Sk4x_sse.h"
     #elif defined(SK_ARM_HAS_NEON)
         #include "Sk4x_neon.h"
@@ skipping 19 matching lines @@
 
     Sk4x(const Sk4x&);
     Sk4x& operator=(const Sk4x&);
 
     static Sk4x Load       (const T[4]);
     static Sk4x LoadAligned(const T[4]);
 
     void store       (T[4]) const;
     void storeAligned(T[4]) const;
 
+    // Experimental!
+    static Sk4x Load2(const T src[2]) {
+        const T padded[4] = { src[0], src[1], 0, 0 };
+        return Load(padded);
+    }
+    void store2(T dst[2]) const {
+        T padded[4];
+        this->store(padded);
+        dst[0] = padded[0];
+        dst[1] = padded[1];
+    }
+
     template <typename Dst> Dst reinterpret() const;
     template <typename Dst> Dst        cast() const;
 
     bool allTrue() const;
     bool anyTrue() const;
 
     Sk4x   bitNot()            const;
     Sk4x   bitAnd(const Sk4x&) const;
     Sk4x    bitOr(const Sk4x&) const;
     // TODO: Sk4x bitAndNot(const Sk4x&) const; is efficient in SSE.
@@ skipping 43 matching lines @@
 
 #if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
     #include "Sk4x_sse.h"
 #elif defined(SK_ARM_HAS_NEON)
     #include "Sk4x_neon.h"
 #else
     #include "Sk4x_portable.h"
 #endif
 
 #endif//Sk4x_DEFINED