[PATCH] libclc/math: Add cospi

Ported from the libclc/amd-builtins branch

v2: Rename sincos_f_piby4 to __libclc__sincosf_piby4
    Add cospi(double) implementation instead of using llvm.cos

The sincosD_piby4.h file is mostly the same as the builtin implementation
released by AMD. The inline attribute declaration is changed, and M_PI is
used instead of a constant double. Otherwise, the only difference is that
the header explicitly enables the fp64 pragma.

Piglit lacks the built-in capability to test double-precision CL builtins
at the moment. I hacked the cospi(float) tests to use cospi(double) to
verify that the implementation runs and produces reasonable results, but
I have NOT added new test cases to piglit which exercise the additional
precision of the double-precision data type for cospi.

Signed-off-by: Aaron Watry <awatry@gmail.com>

Modulo the math, which I cannot really check, this looks good to me.