[PP] feat: optimize PP performance through throttling

[AlpinDale]

Qwen3-30B-A3B, 2x A100, PP=2

Before:

============ Serving Benchmark Result ============
Successful requests:                     2048      
Benchmark duration (s):                  268.78    
Total input tokens:                      1046373   
Total generated tokens:                  1044480   
Request throughput (req/s):              7.62      
Output token throughput (tok/s):         3885.99   
Total Token throughput (tok/s):          7779.01   
---------------Time to First Token----------------
Mean TTFT (ms):                          118972.89 
Median TTFT (ms):                        103316.73 
P99 TTFT (ms):                           241757.32 
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms):                          64.21     
Median TPOT (ms):                        66.32     
P99 TPOT (ms):                           68.40     
---------------Inter-token Latency----------------
Mean ITL (ms):                           64.18     
Median ITL (ms):                         56.77     
P99 ITL (ms):                            188.79    
==================================================

After:

============ Serving Benchmark Result ============
Successful requests:                     2048      
Benchmark duration (s):                  251.71    
Total input tokens:                      1046373   
Total generated tokens:                  1042887   
Request throughput (req/s):              8.14      
Output token throughput (tok/s):         4143.20   
Total Token throughput (tok/s):          8300.26   
---------------Time to First Token----------------
Mean TTFT (ms):                          110023.63 
Median TTFT (ms):                        97427.31  
P99 TTFT (ms):                           223116.10 
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms):                          59.75     
Median TPOT (ms):                        60.11     
P99 TPOT (ms):                           63.72     
---------------Inter-token Latency----------------
Mean ITL (ms):                           59.72     
Median ITL (ms):                         50.97     
P99 ITL (ms):                            173.29    
==================================================