Bridging Language Gaps in Audio-Text Retrieval