Refactor IntegratedLuminosity_cff.py

[lnestor]

Justification

Every time I opened IntegratedLuminosity_cff.py I got annoyed. It is very cluttered, has several helper functions that aren't written well, and isn't clear how to add a new year or new dataset. It also computes the luminosity for every possible combination of eras, and we just don't need that. The was we currently use it mostly is for a single era.

Proposal

I propose a new interface: get_lumi(dataset, year, era). I would also be fine with get_lumi("MET_2017C") for example, just accepting a single string, but I think splitting it makes sense. This new function would calculate just the luminosity requested. That makes this file much simpler to understand. I think it is easy to understand how to add new years and datasets.

I also propose changing the filename to integrated_luminosity.py. I don't care too much about the filename, however the _cff suffix annoys me because this file isn't used in any CMSSW configuration files, and that is what _cff is for. It should have no suffix to be in line with the rest of our codebase. Also, python files should be snake_case.

Backwards Compatibility

Obviously switching the interface from a lumi dictionary to get_lumi will break old scripts. I have added a custom object, that when a script calls lumi["something"] it will raise a deprecation error with a useful error message. This means we don't need to change every part of the code base, we just to incrementally change what we are currently working on. And to be honest, like 50% or more of the scripts in this repository wouldn't even run anyways, because either they are in Python2, the data files no longer exist, etc.

Verification

I have compared every key in the original lumi dictionary to the new get_lumi function, and they all match.