Task #4580: Added template specializations of KernelFactory for FIR_FilterKernel (will probably need some more work). Updated test, it no longer uses external parset file.

Task #4580: Made methods defaultCompileDefintions() and defaultCompileFlags() thread-safe. Cleaned up gpu_utils sources w.r.t. #include's.

Task #4520: Changed order of arguments in createPTX(); definitions are more likely to vary than flags.

Task #4520: Got rid of CudaRuntimeCompiler and KernelCompiler; functionality for this is now in gpu_utils.

* Kernel compilation to PTX can now be done in a new constructor of the
  Kernel base class. 
* Modified createPTX(); added some extra helper functions to gpu_utils.cc
* Fixed a few bugs in createModule() w.r.t. usage of CUjit_options (this is
  still a big mess). 
* TODO: 
  - Removed redundant code
  - Update tests so that they use the new Kernel constructor.

Task #4520: Added CompileFlags class. Moved CompileDefinitions and CompileFlags to KernelCompiler.h file

Task #4520: Moved gpu_utils.h one level up (to the generic directory). Temporarily added method defaultFlags() and typedef of flags_type to this header. Fixed OpenCL code to (temporarily) include OpenCL-specific gpu_utils.h header.

Task #4433: move most of ocl files to opencl/, rename opencl-incl.h -> gpu_incl.h, add OPENCL_ to header guards, complete CMakeLists.txt list of files. Commit now to free up Kernels/, Pipelines/ and WorkQueues/ dir for header fwding headers.

Task #4362: Updated 'global' complex.h file. It now includes either cuda/complex.h or opencl/complex.h (still empty) depending on USE_CUDA and USE_OPENCL.