Switch SkAutoMalloc to SkAutoTMalloc to avoid cast

gm/colorcube.cpp

Index: gm/colorcube.cpp
diff --git a/gm/colorcube.cpp b/gm/colorcube.cpp
index 519f88a6a5e4cfedf2a10b5def6862dbf633e55f..fa9eea9f99098d6b5020070fd233367288ec1ca5 100644
--- a/gm/colorcube.cpp
+++ b/gm/colorcube.cpp
@@ -9,6 +9,7 @@
 #include "SkColorCubeFilter.h"
 #include "SkData.h"
 #include "SkGradientShader.h"
+#include "SkTemplates.h"
 
 namespace skiagm {
 
@@ -72,10 +73,10 @@ protected:
     void make_3Dlut(SkData** data, int size, bool invR, bool invG, bool invB) {
         *data = SkData::NewUninitialized(sizeof(SkColor) * size * size * size);
         SkColor* pixels = (SkColor*)((*data)->writable_data());
-        SkAutoMalloc lutMemory(size);
-        SkAutoMalloc invLutMemory(size);
-        uint8_t* lut = (uint8_t*)lutMemory.get();
-        uint8_t* invLut = (uint8_t*)invLutMemory.get();
+        SkAutoTMalloc<uint8_t> lutMemory(size);
+        SkAutoTMalloc<uint8_t> invLutMemory(size);
+        uint8_t* lut = lutMemory.get();
+        uint8_t* invLut = invLutMemory.get();
         const int maxIndex = size - 1;
         for (int i = 0; i < size; i++) {
             lut[i] = (i * 255) / maxIndex;