Performance plot histogrammar plots deviate from the vanilla pyroot code

[rymuelle]

As Adrian has noted, we see some discrepancies between the two approaches in some plots.

e.g.

vs.

One cause was simply due to using the absolute value of eta in the histogrammar implementation. This is fixed. Some other plots have more subtle issues.

Its worth noting that some plots are exact:

e.g.

Some are similar, but not exact:

And some are wildly different:

[rymuelle]

Ok, so I found and fixed one issue with the mass plots. The selection criteria used the hybrid mass rather than the global mass causing an issue.

Now the plots look quite similar:

But it isn't clear if they are the same. One issue is that since both have a random element there is some noise. I would be ok with accepting some noise, except the vs pt graph looks terrible:

[demarley]

@rymuelle has fixed many of these issues, I'm currently testing it.

[demarley]

From what I can tell, the results are consistent between the two frameworks.
Suggestions:

1. Different marker styles for each type of plot so that you can see them even if they are the same value
2. File naming is inconsistent. For example, the pulls plots are named TH2F_glb_sta_eta_ptPull_gaus_mean.png but the res plots are named glb_sta_eta_v_ptRes_gaus_mean.png.
3. I generally discourage the use of code like this because it obfuscates where functions / variables exist and makes it difficult to repurpose code (IMO the code should be object-oriented -- put in classes -- instead).