Support penalty in overlap mode; return logprob with chunked prefill; improve benchmark scripts (#3988)

Co-authored-by: SangBin Cho <rkooo567@gmail.com>
Co-authored-by: dhou-xai <dhou@x.ai>
Co-authored-by: Hanming Lu <hanming_lu@berkeley.edu>

python/sglang/srt/layers/rotary_embedding.py

@@ -707,7 +707,6 @@ class DeepseekScalingRotaryEmbedding(RotaryEmbedding):
         cos = freqs.cos() * self.mscale
         sin = freqs.sin() * self.mscale
         cache = torch.cat((cos, sin), dim=-1)
-        print("Cache shape", cache.shape)
         return cache
 
     def forward(