fix(PROVEN): _moe_C compiles and runs on real BI-V100 hardware

Tested on real machine (cc-b2042074, BI-V100, IX-ML 3.2.3):
  _moe_C.topk_softmax() → SUCCESS, correct output

Two fixes proven on hardware:
1. cuda_compat.h: WARP_SIZE=64 (BI-V100 warp is 64, not 32)
2. topk_softmax_kernels.cu: cub/block/block_reduce.cuh instead of cub/cub.cuh
   (cub.cuh pulls radix_sort which has WARP_SIZE conflict)

Key finding: ixformer SDK on this base image does NOT have topk_softmax.
The ixformer::infer namespace from xllm's ixformer.h is for newer SDK.
We MUST compile our own _moe_C kernel — which now works.

Build flags (clang 16, ivcore10):
  CUDA: -O3 -cl-fast-relaxed-math (NOT --use_fast_math)
  C++:  -O2 -std=c++17

Dockerfile simplified: 3 steps (was 6)
_custom_ops.py: _moe_C as Priority 0, in-place vllm API

ex_engine/csrc/moe_v055/cuda_compat.h

@@ -5,7 +5,7 @@
 #endif
 
 #ifndef USE_ROCM
-  #define WARP_SIZE 32
+  #define WARP_SIZE 64
 #else
   #define WARP_SIZE warpSize
 #endif

ex_engine/csrc/moe_v055/topk_softmax_kernels.cu

@@ -23,7 +23,7 @@
 
 #ifndef USE_ROCM
     #include <cub/util_type.cuh>
-    #include <cub/cub.cuh>
+    #include <cub/block/block_reduce.cuh>
 #else
     #include <hipcub/util_type.hpp>
     #include <hipcub/hipcub.hpp>