Speed up smart search: both fine and coarse grid

[asmit3]

Moving compute intensive calculations of fine and coarse grid to C++. The C++ code is in rstbx and further optimization will be done there. Some timings on representative LS49 image (timestamp 20180501143628702) For generating timings, see LS49_regression/speedup_smart_search Coarse grid on master--> 82 seconds on KNL Coarse grid on this branch --> 0.72 seconds on KNL Fine grid has similar timings as coarse grids in both branches

[asmit3]

Openmp version gives further speedups Coarse grid with openmp --> 0.47s Fine grid with openmp --> 0.17s