Text this: Cross-Modal Collaboration and Robust Feature Classifier for Open-Vocabulary 3D Object Detection