Introduce general statistics and plotting classes + Addressing comments on the u/v seeding

[bfonta]

This PR addresses all comments made to my first presentation on the new seeding on (U,V) coordinates, namely:

• use Clopper-Pearson intervals for efficiency errors instead of Poisson
• add TCs to efficiency plots but cutting the noise
  • I've applied a cylinder around the center of the shower
• visualize resolution distributions also with violin plots
• calculate the seeding efficiency having as reference the trigger cells around the generated particle, skipping the ROI step. This allows to factor the ROI out of seeding efficiency measurements.

Additionally, the PR includes a new Collector and ChainPlotter classes which will hopefully represent the first step towards unifying the efficiency/resolution statistics collection and plots which are common to all our efforts. If possible, I would very much like most plotting being done through these classes, which would slowly evolve depending on the need of everybody. For instance, I would be happy to add other libraries (besides bokeh). The rationale is to avoid duplicating code when we all do similar plots, such as efficiencies, binned distributions, ...

A new "meta" chain script was added to run individual chains and plots results of multiple chains together, for instance:

Finally, I've added the number of events as an additional command line argument; this ensures all intermediate files stored are different for different number of input events, which is both convenient and safe (no confusion can arise in case one decides to run only part of the chain for a specific study).

As an example, the meta chain can be run as follows:

python bye_splits/scripts/run_meta.py --nevents 4000

Many flags are there for fine-grained running. As an example, to do the data collection and plotting steps a second time without reprocessing the same data:

python bye_splits/scripts/run_meta.py --nevents 4000 --no_roi --no_seed --no_cluster --no_smooth --no_fill

[isehle]

Thanks for all the work you put into this! Besides my one question in cluster.py, this should all be fine for me. Would you also mind putting an example of how we can use the new plotting class in the PR? I.e., a command to run and its output, for example.

[bfonta]

@isehle See scripts/run_meta.py, which should exemplify how to use the two new classes.
Please note that I am aware there are bits in the new classes that are not general, and some adaptations will likely be needed as more people use them and as we understand our needs. If there is something you do not like or think could be improved, now is the moment, since all of us will hopefully be using these from now on!

[mchiusi]

This really looks great! I did fetch your branch to clarify some critical points - especially in the validation - by running the code in local.
The display code in written using bokeh seems at least to me much more convoluted that the same code in other libraries. Anyway I tested and it works.

[bfonta]

@mchiusi I am not enforcing any library, you can use the one you prefer. I would just like that all the code right before plotting is the same for all of us. We can easily add some bokeh/plotly choice later on, as we are already doing in the geometry output.

[bfonta]

I am keeping the three commits for clarity, since they are completely independent.