Only evaluate non-empty responses to reduce eval script latency, cost

Empty responses by Khoj will always be an incorrect response, so no
need to make call to an evaluator agent to check that

tests/eval_frames.py

@@ -113,6 +113,10 @@ def process_batch(batch, counter, results, dataset_length):
         agent_response = get_agent_response(prompt)
 
         # Evaluate response
+        if agent_response is None or agent_response.strip() == "":
+            evaluation["decision"] = False
+            evaluation["explanation"] = "Agent response is empty. This maybe due to a service error."
+        else:
             evaluation = evaluate_response(prompt, agent_response, answer)
 
         # Store results