Parameterize event_type thresholds

[JBernete]

Tarek and I were talking about the binning vs statistics problem: if bins, either in energy or offset, are small, they might not contain enough events to do a good event_type partitioning, but bins too large may reproduce the FOV and energy effects that we want to avoid by binning.
To solve this, Tarek had the idea that if we plot the ET thresholds vs energy or vs offset, there could be a way to smoothly join these points and have the thresholds as a function of the energy and the offset, instead of binned.

We should test if this is possible. And if it is, we should discuss what is the optimal binning to do it. This way maybe we can loosen up a bit the current binning to make sure we have enough statistics in every bin to estimate good thresholds.

[TarekHC]

Hi Juan,

I renamed the issue mainly because I continued thinking about it: Ideally, what we want is to find a nice parameterization for the cuts. Interpolating would mean that we fill cut thresholds over intermediate energies (and most likely, the ones we will need to "guess" are the ones of the highest energies, so probably it would be closer to extrapolating).

In any case, for the moment I believe plotting the logAngDiff vs E for the diffuse gamma test sample (for example, each 1-deg offset separately), and highlight where the thresholds are located, maybe we can come up with a decent spline... Or maybe a linear fit could be enough... who knows.