[Ops][Misc] Optimize split_qkv_rmsnorm_rope op (#6827)

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

This PR optimizes the `split_qkv_rmsnorm_rope` operator by introducing a
new Triton kernel, `split_qkv_rmsnorm_rope_prefill_kernel`, for the
prefill stage (i.e., large batch sizes). The implementation now
dynamically selects between the existing decode kernel and the new
prefill kernel based on the batch size, which improves performance for
large batch scenarios.

Additionally, the RoPE implementation is updated to support partial
rotation dimensions (`rope_dim`), making the operator more flexible.

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

No. This is a performance optimization and is not expected to introduce
any user-facing changes.

### How was this patch tested?

CI should pass with existing tests. The new prefill path is triggered
when the batch size is larger than the number of available vector cores.
The partial RoPE feature can be tested by passing the `rope_dim`
argument.
- vLLM version: v0.15.0
- vLLM main:
83b47f67b1

---------

Signed-off-by: guzhiyong <guzhiyong5@h-partners.com>
Signed-off-by: frank <2547457096@qq.com>
Co-authored-by: guzhiyong <guzhiyong5@h-partners.com>

tests/e2e/singlecard/compile/test_graphex_qknorm_rope_fusion.py

@@ -1,5 +1,6 @@
 import copy
 
+import numpy as np
 import pytest
 import torch
 import torch.nn as nn
@@ -16,6 +17,8 @@ from vllm_ascend.compilation.passes.qknorm_rope_fusion_pass import (
 )
 from vllm_ascend.ops.triton.triton_utils import init_device_properties_triton
 
+MAX_POSITION_EMBEDDING = 262144
+
 
 def find_op(gm, op_default):
     return any(node.op == "call_function" and node.target == op_default for node in gm.graph.nodes)
@@ -207,8 +210,10 @@ def test_rmsnorm_quant_fusion(
         model = model.to("npu")
         seq_len = 5
         qkv = torch.randn(seq_len, qkv_size, device="npu", dtype=dtype)
-        cos = torch.randn(1, seq_len, 1, head_dim, device="npu", dtype=dtype)
-        sin = torch.randn(1, seq_len, 1, head_dim, device="npu", dtype=dtype)
+        cos_sin_cache = torch.from_numpy(np.random.uniform(0, 1, [MAX_POSITION_EMBEDDING, head_dim])).to(dtype).npu()
+        positions = torch.randint(
+            low=0, high=MAX_POSITION_EMBEDDING, size=(num_tokens,), dtype=torch.int64, device="npu"
+        )
 
         with torch.no_grad():
             original_optimize = torchair.npu_fx_compiler._optimize_fx
@@ -218,6 +223,6 @@ def test_rmsnorm_quant_fusion(
 
             compiled_model = torch.compile(model, backend="npugraph_ex", fullgraph=True, dynamic=True)
 
-            compiled_model(qkv, cos, sin)
+            compiled_model(qkv, cos_sin_cache, positions)
 
             torchair.npu_fx_compiler._optimize_fx = original_optimize