Fix precedence bug in dict-based reasoning extraction

Address gemini-code-assist review feedback:
- Use two-argument dict.get() to check key presence instead of truthiness
- This ensures reasoning_content takes precedence even with falsy values
- Update test to expect empty string instead of None for consistency

The previous 'or' operator would incorrectly fall back to 'reasoning'
when 'reasoning_content' was present but had a falsy value like ''.
Now using dict.get(key, default) properly maintains precedence.

Author: mikkokirvesoja

src/google/adk/models/lite_llm.py

@@ -157,7 +157,7 @@ def _extract_reasoning_value(message: Message | Dict[str, Any]) -> Any:
     return getattr(message, "reasoning")
 
   if isinstance(message, dict):
-    return message.get("reasoning_content") or message.get("reasoning")
+    return message.get("reasoning_content", message.get("reasoning"))
 
   return None
 

tests/unittests/models/test_litellm.py

@@ -1706,9 +1706,8 @@ def test_extract_reasoning_value_with_empty_reasoning_content():
   }
 
   result = _extract_reasoning_value(message)
-  # Empty string from reasoning_content is returned, but 'or' makes it None
-  # because empty string is falsy and reasoning is not present
-  assert result is None
+  # Empty string should be returned to maintain precedence
+  assert result == ""
 
 
 def test_parse_tool_calls_from_text_multiple_calls():