Reset changes to minimum necessary to implement our PSD sampling

In an attempt to try to limit the divergence from the master bilby, I've scrapped most of the changes to the likelihood and interferometer modules. The likelihood in this MR should correctly account for the normalization factor. I've also fixed the prob method of the UniformIntegers prior class, and set the maximum frequency of the interferometers in the example to 1023.75 to avoid having the last index be infinite. I also made the method to load the files from BayesWave more efficient.