Computational Complexity of Detecting Proximity to Losslessly Compressible Neural Network Parameters