Fix reduction and norm register spill

### 🐛 Describe the bug

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIN3c104HalfENS1_9ArgMinOpsIfEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 94

Kernel: typeinfo name for at::native::xpu::ReduceKernel<2, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi2ENS1_8ReduceOpIsNS1_9ArgMinOpsIsEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 88

Kernel: typeinfo name for at::native::xpu::ReduceKernel<2, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi2ENS1_8ReduceOpIiNS1_9ArgMinOpsIiEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 84

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIfNS1_9ArgMinOpsIfEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 143

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIlNS1_9ArgMinOpsIlEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 131

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIiNS1_9ArgMinOpsIiEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 31

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIN3c108BFloat16ENS1_9ArgMinOpsIfEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 87

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIN3c104HalfENS1_9ArgMaxOpsIfEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 94

Kernel: typeinfo name for at::native::xpu::ReduceKernel<2, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi2ENS1_8ReduceOpIsNS1_9ArgMaxOpsIsEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 88

Kernel: typeinfo name for at::native::xpu::ReduceKernel<2, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi2ENS1_8ReduceOpIiNS1_9ArgMaxOpsIiEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 84

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIfNS1_9ArgMaxOpsIfEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 143

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIlNS1_9ArgMaxOpsIlEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 131

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIiNS1_9ArgMaxOpsIiEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 31

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, long, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIN3c108BFloat16ENS1_9ArgMaxOpsIfEEjlLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 87

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, c10::Half, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIN3c104HalfENS1_6MinOpsIS5_EEjS5_Li4EEEEE compiled SIMD32 allocated 256 regs and spilled around 83

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, bool, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIbNS1_6MinOpsIbEEjbLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 104

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, float, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIfNS1_6MinOpsIfEEjfLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 156

Kernel: typeinfo name for at::native::xpu::ReduceKernel<2, at::native::xpu::ReduceOp, unsigned int, int, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi2ENS1_8ReduceOpIiNS1_6MinOpsIiEEjiLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 3

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, int, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIiNS1_6MinOpsIiEEjiLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 135

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, c10::Half, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIN3c104HalfENS1_6MaxOpsIS5_EEjS5_Li4EEEEE compiled SIMD32 allocated 256 regs and spilled around 83

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, bool, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIbNS1_6MaxOpsIbEEjbLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 108

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, float, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIfNS1_6MaxOpsIfEEjfLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 156

Kernel: typeinfo name for at::native::xpu::ReduceKernel<2, at::native::xpu::ReduceOp, unsigned int, int, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi2ENS1_8ReduceOpIiNS1_6MaxOpsIiEEjiLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 3

Kernel: typeinfo name for at::native::xpu::ReduceKernel<4, at::native::xpu::ReduceOp, unsigned int, int, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi4ENS1_8ReduceOpIiNS1_6MaxOpsIiEEjiLi4EEEEE compiled SIMD32 allocated 256 regs and spilled around 135

Kernel: typeinfo name for at::native::xpu::ReduceKernel<1, at::native::xpu::ReduceOp, unsigned int, float, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi1ENS1_8ReduceOpIN3c104HalfENS1_7NormOpsIS5_ffEEjfLi4EEEEE compiled SIMD32 allocated 128 regs and spilled around 12

Kernel: typeinfo name for at::native::xpu::ReduceKernel<2, at::native::xpu::ReduceOp, unsigned int, float, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi2ENS1_8ReduceOpIN3c104HalfENS1_7NormOpsIS5_ffEEjfLi4EEEEE compiled SIMD32 allocated 128 regs and spilled around 12

Kernel: typeinfo name for at::native::xpu::ReduceKernel<1, at::native::xpu::ReduceOp, unsigned int, c10::Half, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi1ENS1_8ReduceOpIN3c104HalfENS1_7NormOpsIS5_fS5_EEjS5_Li4EEEEE compiled SIMD32 allocated 128 regs and spilled around 12

Kernel: typeinfo name for at::native::xpu::ReduceKernel<2, at::native::xpu::ReduceOp, unsigned int, c10::Half, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi2ENS1_8ReduceOpIN3c104HalfENS1_7NormOpsIS5_fS5_EEjS5_Li4EEEEE compiled SIMD32 allocated 128 regs and spilled around 12

Kernel: typeinfo name for at::native::xpu::ReduceKernel<1, at::native::xpu::ReduceOp, unsigned int, float, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi1ENS1_8ReduceOpIN3c108BFloat16ENS1_7NormOpsIS5_ffEEjfLi4EEEEE compiled SIMD32 allocated 128 regs and spilled around 12

Kernel: typeinfo name for at::native::xpu::ReduceKernel<2, at::native::xpu::ReduceOp, unsigned int, float, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi2ENS1_8ReduceOpIN3c108BFloat16ENS1_7NormOpsIS5_ffEEjfLi4EEEEE compiled SIMD32 allocated 128 regs and spilled around 12

Kernel: typeinfo name for at::native::xpu::ReduceKernel<1, at::native::xpu::ReduceOp, unsigned int, c10::BFloat16, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi1ENS1_8ReduceOpIN3c108BFloat16ENS1_7NormOpsIS5_fS5_EEjS5_Li4EEEEE compiled SIMD32 allocated 128 regs and spilled around 12

Kernel: typeinfo name for at::native::xpu::ReduceKernel<2, at::native::xpu::ReduceOp, unsigned int, c10::BFloat16, 4> >
Warning: warning: kernel _ZTSN2at6native3xpu12ReduceKernelILi2ENS1_8ReduceOpIN3c108BFloat16ENS1_7NormOpsIS5_fS5_EEjS5_Li4EEEEE compiled SIMD32 allocated 128 regs and spilled around 12

### Versions

main