fix: Update eval metric according to the new vertexai eval package.

llm_demo/evaluation/evaluation.py

@@ -68,7 +68,7 @@ def evaluate_retrieval_phase(
     """
     Run evaluation for the ability of a model to select the right tool and arguments (retrieval phase).
     """
-    metrics = ["tool_call_quality"]
+    metrics = ["tool_call_valid"]
     # Prepare evaluation task input
     responses = []
     references = []