[BUG] Fail to build on A3

[raintBN-91]

I have already try this repo on A2 with no modifications needed.

But when build on A3, build.sh will report the below bug:

[ 97%] Building CXX object csrc/CMakeFiles/sgl_kernel_npu.dir/pytorch_extensions.cpp.o
In file included from /home/jiaozeyu/tmp/sgl-kernel-npu/csrc/cache_location_assign/op_host/cache_loc_assign.cpp:13:
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/utils/common.h:15:11: error: ‘uint32_t’ does not name a type
   15 | constexpr uint32_t BLK_SIZE_ALIN_FOR_INT64 = 4;
      |           ^~~~~~~~
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/utils/common.h:1:1: note: ‘uint32_t’ is defined in header ‘<cstdint>’; did you forget to ‘#include <cstdint>’?
  +++ |+#include <cstdint>
    1 | 
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/utils/common.h:16:11: error: ‘uint32_t’ does not name a type
   16 | constexpr uint32_t BLK_SIZE_ALIN_FOR_INT32 = 8;
      |           ^~~~~~~~
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/utils/common.h:16:11: note: ‘uint32_t’ is defined in header ‘<cstdint>’; did you forget to ‘#include <cstdint>’?
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/utils/common.h:18:8: error: ‘uint64_t’ does not name a type
   18 | inline uint64_t alinInt64Count(uint64_t count)
      |        ^~~~~~~~
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/utils/common.h:18:8: note: ‘uint64_t’ is defined in header ‘<cstdint>’; did you forget to ‘#include <cstdint>’?
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/utils/common.h:23:8: error: ‘uint64_t’ does not name a type
   23 | inline uint64_t alinInt32Count(uint64_t count)
      |        ^~~~~~~~
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/utils/common.h:23:8: note: ‘uint64_t’ is defined in header ‘<cstdint>’; did you forget to ‘#include <cstdint>’?
[ 98%] Linking CXX shared module deep_ep_cpp.cpython-311-aarch64-linux-gnu.so

Finally result in:

/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/mla_preprocess/op_host/mla_preprocess.cpp:685:31: warning: comparison of integer expressions of different signedness: ‘int32_t’ {aka ‘int’} and ‘const uint32_t’ {aka ‘const unsigned int’} [-Wsign-compare]
  685 |     if (bIndex >= 0 && bIndex < MAX_SUPPORT_TOKEN_NUMS) {
      |                        ~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/cache_location_assign/op_host/cache_loc_assign.cpp: In function ‘at::Tensor sglang::npu_kernel::getTiling(const at::Tensor&, uint64_t, uint64_t, uint32_t&, bool)’:
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/cache_location_assign/op_host/cache_loc_assign.cpp:47:42: error: ‘alinInt32Count’ was not declared in this scope
   47 |         tillingData->reqInxBufferCount = alinInt32Count(batchSize);
      |                                          ^~~~~~~~~~~~~~
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/cache_location_assign/op_host/cache_loc_assign.cpp:51:42: error: ‘alinInt64Count’ was not declared in this scope
   51 |         tillingData->reqInxBufferCount = alinInt64Count(batchSize);
      |                                          ^~~~~~~~~~~~~~
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/cache_location_assign/op_host/cache_loc_assign.cpp:55:41: error: ‘alinInt32Count’ was not declared in this scope
   55 |     tillingData->tokenCountAlignInt32 = alinInt32Count(MAX_STEP);
      |                                         ^~~~~~~~~~~~~~
/home/jiaozeyu/tmp/sgl-kernel-npu/csrc/cache_location_assign/op_host/cache_loc_assign.cpp:58:42: error: ‘alinInt64Count’ was not declared in this scope
   58 |     tillingData->offsetCountAlignInt64 = alinInt64Count(batchSize);
      |                                          ^~~~~~~~~~~~~~
gmake[2]: *** [csrc/CMakeFiles/sgl_kernel_npu.dir/build.make:107: csrc/CMakeFiles/sgl_kernel_npu.dir/cache_location_assign/op_host/cache_loc_assign.cpp.o] Error 1
gmake[2]: *** Waiting for unfinished jobs....
gmake[1]: *** [CMakeFiles/Makefile2:766: csrc/CMakeFiles/sgl_kernel_npu.dir/all] Error 2
gmake: *** [Makefile:136: all] Error 2

And only "torch_memory_saver-0.0.8-cp311-cp311-linux_aarch64.whl" will be built.