Update README.md

powdist.m defines a function which briefly analyses a light curve to search for crosstalk signatures. 

It plots the light curve together with a simple Leahy-normalized power spectrum (PS) of the data and the same PS rebinned onto a quasi-logarithmic grid. Then, assuming that the observation is mostly composed of Poissonian noise modified by crosstalk, it generates the expected distribution in three different crosstalk models, and plots the counts-per-bin distribution both in the data and the models.

Plots are produced both with and without the `-nodesktop` option, but in the second case an interactive window will open too.

By checking that the high-frequency PSs are flat (i.e., the noise tends to white noise), and that the alignment between at least one of the models and the data count distribution is good, users can safely just renormalize the powers in their PSs then by dividing them by half the average power at high frequencies.

The function takes five arguments as input  (NyqFreq,LC,Binned,PSname,Noisename) and returns three (extrVoverE,CrossTalkProb,HalfMeanPow). \

Input arguments:

- `CrossTalkProb` is the total crosstalk probability (it only assumes underlying poissonian noise and is independent of the number of pixels that can be fired by a single photon)

- `HalfMeanPow` is half the average power at frequencies higher than 1000 Hz in the power spectrum (this is what extrVoverE has to be compared to)

Applying this simple analysis will inevitably produce dubious results when used on data with gaps in it, and data with a strong presence of high-fluence bursts: it is for this reason that double-checking the light curve is necessary whenever the modified distributions do not agree with the data.

<details><summary>Sample graphs from powdist</summary>

For reference, we show here the two figures produced by powdist running on the first 512 seconds of the SiFAP2 3FGLJ1544_20180418_N1_Bary.fits observation with the command `powdist(Nyq,tsplit,false,'grafiPS.png','grafiCT.png')`. \