Update to improve performance of SimpleFlow algorithm

+ Improve performance of calcOpticalFlowSingleScale method
+ Small refactoring

Current results:

IMAGE NAMES          RMSE
Beanbags
Dimetrodon           0.329428
DogDance
Grove2               0.550852
Grove3               1.464699
Hydrangea            0.523277
MiniCooper
RubberWhale          0.367246
Urban2               2.717003
Urban3               4.185070
Venus                0.775422
Walking

Time (for Urban3):
17.490248 sec

modules/video/src/simpleflow.hpp

@@ -51,6 +51,22 @@ using namespace std;
 #define UNKNOWN_FLOW_THRESH 1e9
 
 namespace cv {
+/*
+template<class T>
+inline static T sqr(T t) {
+  return t*t;
+}
+
+static float dist(const Vec3b& p1, const Vec3b& p2) {
+  return sqr(p1[0] - p2[0]) +
+         sqr(p1[1] - p2[1]) +
+         sqr(p1[2] - p2[2]);
+}
+
+inline static float dist(const Vec2f& p1, const Vec2f& p2) {
+  return sqr(p1[0] - p2[0]) +
+         sqr(p1[1] - p2[1]);
+}*/
 
 inline static float dist(const Vec3b& p1, const Vec3b& p2) {
   return (p1[0] - p2[0]) * (p1[0] - p2[0]) +