Removed redundant copy ut track loops and removed unused variables in UT search window.

1. Each thread in ut_copy_track_hit_number loops over all the velo-UT tracks in the event with increments of 1, when it can just increment by blockDim.x to avoid redundant copies.
2. UT search windows copies fudge factors for all UT layers, but each thread only looks for search window in 1 UT layer and only need 1 fudge factor.
3. Exact same physics results in tests.

device/UT/compassUT/src/SearchWindows.cu

@@ -167,17 +167,15 @@ __device__ std::tuple<int, int, int, int, int, int, int, int, int, int> calculat
   const float min_pt,
   const float min_momentum)
 {
-  // TODO: Understand and fix this logic
+  // TODO: Understand and fix this logic (especially ty_index)
   // -- This is hardcoded, so faster
   // -- If you ever change the Table in the magnet tool, this will be wrong
+  // -- Need to understand where 100 and 0.5 comes from when converting ty->ty_index
   const float absSlopeY = fabsf(velo_state.ty);
-  int index = static_cast<int>(absSlopeY * 100 + 0.5f);
-  // assert(3 + 4 * index < UTMagnetTool::N_dxLay_vals);
-  if (3 + 4 * index >= UTMagnetTool::N_dxLay_vals) {
-    index = (UTMagnetTool::N_dxLay_vals - 3) / 4 - 1;
-  }
-  const float normFact[4] {
-    fudge_factors[4 * index], fudge_factors[1 + 4 * index], fudge_factors[2 + 4 * index], fudge_factors[3 + 4 * index]};
+  const int ty_index = static_cast<int>(absSlopeY * 100 + 0.5f);
+  const int fudge_index =
+    min(UT::Constants::n_layers * ty_index + layer, UTMagnetTool::N_dxLay_vals - UT::Constants::n_layers + layer);
+  const float normFact = fudge_factors[fudge_index];
 
   // -- this 500 seems a little odd...
   // to do: change back!
@@ -190,7 +188,7 @@ __device__ std::tuple<int, int, int, int, int, int, int, int, int, int> calculat
   const float z_at_layer = ut_hits.zAtYEq0(layer_offset);
   const float y_track = velo_state.y + velo_state.ty * (z_at_layer - velo_state.z);
   const float x_track = velo_state.x + velo_state.tx * (z_at_layer - velo_state.z);
-  const float invNormFact = 1.0f / normFact[layer];
+  const float invNormFact = 1.0f / normFact;
   const float xTolNormFact = xTol * invNormFact;
 
   // Find sector group for lowerBoundX and upperBoundX