Removing 32 bit Mac support

With Apple moving to clang as the default compiler in 10.8 and the removal of gcc in Xcode 5 (though I believe you can get gcc back by installing older versions of the tools), we have to take a hard look at 32 bit support.

When clang is used as the host compiler, 32 bit target compilation is not supported. This is because clang doesn’t support the -malign-double switch nvcc needs to properly align double-precision structure fields (gcc supports this flag).

http://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html#errata-cuda-tools-5-5

Since support for building 32 bits is going away, we will need to remove the universal binary support currently implemented in OptiX and limit it to 64 bits.

Will any of you be greatly affected by this?