Fix bug when using conversation with async call

Refs: simonw/llm#674
---------

Co-authored-by: Sukhbinder Singh <sukhbindersingh@gmail.com>

llm_gemini.py

@@ -155,7 +155,7 @@ def build_messages(self, prompt, conversation):
                 if response.prompt.prompt:
                     parts.append({"text": response.prompt.prompt})
                 messages.append({"role": "user", "parts": parts})
-                messages.append({"role": "model", "parts": [{"text": response.text()}]})
+                messages.append({"role": "model", "parts": [{"text": response.text_or_raise()}]})
 
         parts = []
         if prompt.prompt:
@@ -335,4 +335,4 @@ def embed_batch(self, items):
             )
 
         response.raise_for_status()
-        return [item["values"] for item in response.json()["embeddings"]]
+        return [item["values"] for item in response.json()["embeddings"]]