boost paralal of graphrag (#1924)

### What problem does this PR solve?


### Type of change

- [x] Performance Improvement

graphrag/index.py

 
     assert left_token_count > 0, f"The LLM context length({llm_bdl.max_length}) is smaller than prompt({ext.prompt_token_count})"
 
-    BATCH_SIZE=1
+    BATCH_SIZE=4
     texts, graphs = [], []
     cnt = 0
     threads = []
-    exe = ThreadPoolExecutor(max_workers=12)
+    exe = ThreadPoolExecutor(max_workers=50)
     for i in range(len(chunks)):
         tkn_cnt = num_tokens_from_string(chunks[i])
         if cnt+tkn_cnt >= left_token_count and texts: