Simple, AMReX independent test to explore building a CUDA code
either as an executable or through a static library in the AMReX fashion.

This version is the pure C++ build test.

Files
----- 
main.cpp contains the main which calls two CUDA global functions.
MyKernel.cpp and MyKernelB.cpp contain the CUDA global functions.

Make targets
------------
make: Build an executable directly using nvcc. Used to test the code.
make pgi: Build an executable directly using pgfortran, similar to AMReX.

make ***exec: Build an executable through a static library. 
make ***library: Just build the static library.

*** options: pgi, gcc, nvcc
