[x86] Stub out the integer lowering of 256-bit vectors with AVX2

support. No interesting functionality yet, but this will let me
implement one vector type at a time.

llvm-svn: 218277