fix: token limit inversion (#923)

# Motivation

<!-- Why is this change necessary? -->

# Content

<!-- Please include a summary of the change -->

# Testing

<!-- How was the change tested? -->

# Please check the following before marking your PR as ready for review

- [ ] I have added tests for my changes
- [ ] I have updated the documentation or added new documentation as
needed

Author: jayhack

src/codegen/extensions/tools/observation.py

@@ -56,7 +56,7 @@ def render_as_string(self, max_tokens: int = 8000) -> str:
         their string output format.
         """
         rendered = json.dumps(self.model_dump(), indent=2)
-        if 3 * len(rendered) > max_tokens:
+        if len(rendered) > (max_tokens * 3):
             logger.error(f"Observation is too long to render: {len(rendered) * 3} tokens")
             return rendered[:max_tokens] + "\n\n...truncated...\n\n"
         return rendered