[matching] Fix a typo in the preprocessor used to detect SSE support

- This L2 distance specialization is only used for floating point based descriptors.
- So the impact will be visible only if you are using AKAZE_FLOAT.
- Fast experiments show that SSE distance can be up to 40% faster.

src/openMVG/matching/metric.hpp

@@ -69,7 +69,7 @@ struct L2_Vectorized
   }
 };
 
-#ifdef OPENVMG_USE_SSE
+#ifdef OPENMVG_USE_SSE
 
 namespace optim_ss2{
 
@@ -83,7 +83,7 @@ namespace optim_ss2{
   };
 
   // Euclidean distance (SSE method) (squared result)
-  inline float l2_sse(float * b1, float * b2, int size)
+  inline float l2_sse(const float * b1, const float * b2, int size)
   {
     float* b1Pt = (float*)b1;
     float* b2Pt = (float*)b2;
@@ -131,7 +131,7 @@ struct L2_Vectorized<float>
   }
 };
 
-#endif // OPENVMG_USE_SSE
+#endif // OPENMVG_USE_SSE
 
 }  // namespace matching
 }  // namespace openMVG