Can time‐to‐detection models with fewer survey replicates provide a robust alternative to traditional site‐occupancy models?