Use apply_rope_with_cos_sin_cache_inplace for DeepSeek (#4764)

Co-authored-by: Zhang Kaihong <zhangkaihong.zkh@alibaba-inc.com>

python/sglang/srt/layers/rotary_embedding.py

@@ -645,7 +645,7 @@ class DeepseekScalingRotaryEmbedding(RotaryEmbedding):
         cache = torch.cat((cos, sin), dim=-1)
         return cache
 
-    def forward(
+    def forward_native(
         self,
         positions: torch.Tensor,
         query: torch.Tensor,