[CUDA] Do not merge CUDA target attributes.

CUDA target attributes are used for function overloading and must not be merged.

This fixes a bug where attributes were inherited during function template
specialization in CUDA and made it impossible for specialized function
to provide its own target attributes.

Differential Revision: https://reviews.llvm.org/D24522

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@281406 91177308-0d34-0410-b5e6-96231b3b80d8