Lomb-Scargle can not compute window power spectrum
The current implementation to compute the Lomb-Scargle periodogram does not allow to compute the window power spectrum (that is, the power spectrum for a given measured time intervals with counts all = 1).
The reason is that the function 'scargle' subtracts the mean of the signal, which renders the result useless (NaN), or all 0 (in case of the "old" qualifier).
Would it be better to make this a qualifier? And in which direction? Allways subtract mean unless qualifier is given?