Don't assert the exact word in the LiteLLM smoke test

The test checks that a request reaches the executor model through
LiteLLM and text comes back. It also asserted the model obeyed "Reply
with exactly one word: pong", which Haiku once answered with "ping" and
failed CI on #66. Assert non-empty text instead.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE

tests/integration/test_llm_live.py

@@ -26,7 +26,10 @@ def test_executor_model_completes_via_litellm(
     result = litellm.complete(events, tools=None, model=live_config.roles.executor)
     texts = [e.text for e in result if isinstance(e, AssistantText)]
     assert texts, f"no assistant text in {result!r}"
-    assert "pong" in texts[0].lower()
+    # A plumbing check, so any answer will do: what the model chooses to say
+    # is not under test, and asserting the word made CI flaky (it once
+    # replied "ping").
+    assert texts[0].strip(), f"empty assistant text in {result!r}"
 
 
 def test_declared_limits_reach_a_budgeted_request(