AMDGPU: Fix not handling atomicrmw fadd in exotic address spaces correctly

We try to interpret unknown address space numbers as aliases of global,
but this wasn't applied here. Also improve test coverage for the
buffer fat pointer address space.

Author: arsenm

llvm/lib/Target/AMDGPU/SIISelLowering.cpp

@@ -16062,7 +16062,7 @@ SITargetLowering::shouldExpandAtomicRMWInIR(AtomicRMWInst *RMW) const {
       if (HasSystemScope)
         return AtomicExpansionKind::CmpXChg;
 
-      if ((AS == AMDGPUAS::GLOBAL_ADDRESS ||
+      if ((AMDGPU::isExtendedGlobalAddrSpace(AS) ||
            AS == AMDGPUAS::BUFFER_FAT_POINTER) &&
           Ty->isFloatTy()) {
         // global/buffer atomic fadd f32 no-rtn: gfx908, gfx90a, gfx940, gfx11+.