Add normalization of PSD to documentation

[dhuppenkothen]

I'm currently poking around in some data using gatspy, and wanted to take a stab at calculating some false-alarm probabilities. It took me a while of reading the gatspy source code to figure out that the score method returns the PSD in the least-squares normalization (?). Perhaps this is obvious to anyone with more experience with Lomb-Scargle periodograms than me, but for newcomers, it might be useful to explicitly mention that in the docs?

I'm happy to submit a PR, but I wanted to make sure that I'd understood the correctly what it returns before I do that. :)

[jakevdp]

For false alarm probabilities, I'd use astropy. That implementation is much further developed than this one: http://docs.astropy.org/en/stable/stats/lombscargle.html#peak-significance-and-false-alarm-probabilities

[jakevdp]

Also the treatment of normalizations there is much more consistent, and it can compute false alarm scores for all available normalizations.