Give get_intensity() an option to use statistics of time-integrated spectrum

[david-macmahon]

get_intensity uses the statistics of the non-integrated frame to compute the intensity for a desired SNR. The involves dividing by the square root of the number of time samples in the frame, presumably to account for the sqrt(N) increase in the noise after integrating the time samples together. This is valid for purely noise-like signals, but for signals from "real" data the change in SNR after integration can be substantially different. The use of sigma clipping can further exacerbate this. The attached histograms of a frame real data from GBT shows how the standard deviation of the sigma trimmed pre- and post-integrated data can vary substantially (over 3x in this case).

Giving get_intensity an option to use the statistics of the time-integrated spectrum instead of the statistics of the flattened (and non-integrated) frame would be a welcomed addition.