Parameters of PGD and optimal weighting/cic compensation

Hi @modichirag , I compared the matter power spectrum computed with and without PDG. I need know what do you think about these results!
Initially I used kl, ks, mu, and alpha0 parameters obtained by running this [Notebook](https://github.com/VMBoehm/MADLens/blob/3d024515098bb591e5c7f7b4e34b7e1903f8c3ab/notebooks/PGD_params.ipynb) with the following setting for the N-Body simulation :
```box_size=128.
nc = 128
field_size = 5.
nsteps=40
B=2
```
This is how the power spectrum looks like for all the snapshots:

Because of the alpha parameter we used for the PGD is scale factor dependent according the following relation:
`alpha=alpha0*a**mu`,
I tried to see how the amplitude of the power spectrum at small scales changes as a function of alpha0 parameter:
I see that the power spectra that better approximate the analytical one are the power spectra obtained with alpha0=0.008, alpha0=0.009, alpha0=0.01, for bigger and smaller values they start to lose power.
This is how the power spectrum looks like for all the snapshots using alpha0=0.01:
Now, I think that probably my scale translation is not right. The kl,ks obtained with the Vanessa's notebook should be in hMpc^(-1), but I'm not sure.
So, still using the the alpha0 obtained from the notebook (alpha0=0.00269), I tried different scale translation:
```
kl_array=[kl*0.7*0.5/(nc*B/ box_size),kl*0.5/(nc*B/ box_size),kl*2/(nc*B/ box_size),kl*2*0.7/(nc*B/ box_size) ]
ks_array=[ks*0.7*0.5/(nc*B/ box_size),ks*0.5/(nc*B/ box_size),ks*2/(nc*B/ box_size),ks*2*0.7/(nc*B/ box_size)]
```
Here the peaks of the PGD kernel :
and the power spectrum of only one snapshot:
From this plot looks like the right scaling factor for kl and ks are obtained multiplying them for h and 0.5 (this depends how we defined k) .
Assuming this is the better rescaling, I used it to see again how the power spectrum at small scales changes as a function of alpha0
And, alpha0=0.01 still seems like the better choice.