Full-Frequency Temporal Patching and Structured Masking for Enhanced Audio Classification