SQuTR: A Robustness Benchmark for Spoken Query to Text Retrieval under Acoustic Noise