add mla_preprocess kernel (#3226)

### What this PR does / why we need it?

- Adds the `mla_preprocess` custom kernel to provide an optimized
pre-processing operator for Multi-head Latent Attention (MLA) on Ascend
NPUs.
- Wires the new kernel into the C++ extension pipeline so vLLM can
invoke it directly, cutting Python-side tensor shuffling and memory
copies that previously bottlenecked MLA compilation paths.

### Does this PR introduce any user-facing change?

- No. The change only introduces a low-level kernel; public APIs and
inference behavior remain unchanged.

### How was this patch tested?

- Dedicated Ascend kernels are not covered by our CI yet, so no extra
automated tests were added. Future MLA-focused regression runs will
cover this path.

- vLLM version: v0.11.0

Signed-off-by: Chen Chen <0109chenchen@gmail.com>

csrc/ops.h

@@ -124,4 +124,40 @@ namespace vllm_ascend {
         uint32_t output_hidden_dim,
         uint32_t slice_offset,
         uint32_t output_full_dim);
+
+    extern void mla_preprocess_impl(
+        void* stream,
+        void* hidden_state,
+        void* gamma1,
+        void* beta1,
+        void* quant_scale1,
+        void* quant_offset1,
+        void* wdqkv,
+        void* bias1,
+        void* gamma2,
+        void* beta2,
+        void* quant_scale2,
+        void* quant_offset2,
+        void* gamma3,
+        void* sin1,
+        void* cos1,
+        void* sin2,
+        void* cos2,
+        void* keycache,
+        void* slot_mapping,
+        void* wuq,
+        void* bias2,
+        void* wuk,
+        void* descale1,
+        void* descale2,
+        void* ctkv_scale,
+        void* qnope_scale,
+        void* q,
+        void* keycache_out,
+        void* q2,
+        void* keycache_out2,
+        void* workspace,
+        void* tiling,
+        const uint32_t block_dim
+    );
 }