oops, clean up

Complete re-write, update examples and README.

Add a paperscript file for generating Whittle-type MLEs that can then in
turn be used as estimators in a true maximum likelihood computation.

Adding an optional 'longmemory' flag for making sure that the frequency
grid of the FFT doesn't contain exactly zero.

Paper release, including scripts used for the bulk of the model fitting.

More complete example files, now with everything necessary to simulate
and fit data with this model.

Populating commit