Generalize `GaussianCDFEncoding` to arbitrary CDF encoding

This is some generality improvements for the current GaussianCDFEncoding and Dispersion. In general, any CDF could be used in the source code of the encoding; one could store the CDF function in the encoding struct. E.g., give some timeseries x generate the function:

m, s = mean(x), std(x)
f = x -> gaussian(x, m, s)

Any other univariate function instead of f could be generated. This function then is stored as a field in a new struct CDFEncoding, that uses the exact source code of GaussianCDFEncoding but using f instead of the existing gaussian function.

Then, this is super easily propagated into Dispersion: that type should initrialize a CDFEncoding and store the encoding directly as its field. If given only a timeseries, it defaults to getting mean, std and initializing the Gaussian encoding.