build_generator: llm_agent: reduce sleep time (#1024)

I found 60 seconds to be quite excessive and makes experimentation
really slow. I did not find any problems with a reduced sleep time.

Author: DavidKorczynski

experimental/build_generator/llm_agent.py

@@ -247,8 +247,8 @@ def execute(self, result_history: list[Result]) -> BuildResult:
     try:
       client = self.llm.get_chat_client(model=self.llm.get_model())
       while prompt:
-        # Sleep for a minute to avoid over RPM
-        time.sleep(60)
+        # Sleep shortly to avoid RPM
+        time.sleep(6)
 
         response = self.chat_llm(cur_round,
                                  client=client,