ComputeCPP enabled

This version went through the refactoring to allow the code to build with Codeplay's ComputeCPP.
The primary changes were:

• ComputeCPP did not like member function template specializations in class scope. So I replaced the permude templates with explicit permuteX, permuteY, permuteZ and permuteT functions.
• ComputeCPP seemed to not like OpenMP so I had to put a guard around the OpenMP Include header.

Updated Readme

Updates release v0.0.1 by improving the README file.

Initial Release

This release features code which compiles, and when the tests run, they pass.
There is a large set of tests, but the main ones are test/test_dsalsh_sycl and test/test_dslash_sycl_vperf.

In addition:

• complex numbers are implemented as either Fortran like (RIRIRIRI) or as separated real and imaginary parts (more vector oriented) (RRRRIIII)

• Code has compiled with the Intel LLVM/SyCL compiler (see the repository github.com:bjoo/SyCLDslashWorkspace for examples ) but no other compilers were tried so far (e.g. Codeplay)

Known issues:
*) With my current compilations the OpenCL runtime occasionally throws the following
OpenCL API failed. OpenCL API returns: -34 (CL_INVALID_CONTEXT)
which I guess is the OpenCL equivalent of a crash...