[BACKEND] Add arith::CeilFloorDivExpandOpsPatterns (#5934)

Unfortunately `CeilFloorDivExpandOpsPatterns` is not included in
`mlir::arith::populateArithToLLVMConversionPatterns` but still included
in `populateConvertToLLVMConversionPatterns`.


https://github.com/llvm/llvm-project/blob/main/mlir/lib/Conversion/ArithToLLVM/ArithToLLVM.cpp#L544

third_party/nvidia/lib/TritonNVIDIAGPUToLLVM/TritonGPUToLLVM.cpp

@@ -6,6 +6,7 @@
 #include "mlir/Conversion/GPUToNVVM/GPUToNVVMPass.h"
 #include "mlir/Conversion/MathToLLVM/MathToLLVM.h"
 #include "mlir/Conversion/UBToLLVM/UBToLLVM.h"
+#include "mlir/Dialect/Arith/Transforms/Passes.h"
 #include "mlir/Dialect/LLVMIR/LLVMDialect.h"
 #include "mlir/Dialect/LLVMIR/NVVMDialect.h"
 #include "mlir/Pass/Pass.h"
@@ -141,6 +142,7 @@ struct ConvertTritonGPUToLLVM
     // TODO(thomas): this should probably be done in a separate step to not
     // interfere with our own lowering of arith ops. Add arith/math's patterns
     // to help convert scalar expression to LLVM.
+    mlir::arith::populateCeilFloorDivExpandOpsPatterns(patterns);
     mlir::arith::populateArithToLLVMConversionPatterns(typeConverter, patterns);
     mlir::populateMathToLLVMConversionPatterns(typeConverter, patterns);
     mlir::populateGpuToNVVMConversionPatterns(typeConverter, patterns);