runner re-styling and other fixes

[mlizzo]

This PR is meant to apply some resyling to the mkShapesRDF runner and other fixes:

1. splitSamples is internally called by RunAnalysis, such as inputs are the python dictionaries defined in a given configuration
2. script.py contains the samples dictionary rather than the splitSample tuple
3. aliases, samples and nuisances files are updated on-the-fly, such as subsamples are correctly propragated. Now it is possible to define aliases and nuisances per subsample; if only the parent name is given, they will be extended to each subsample - if any
4. runner.py loads all aliases at once and all systematics (shape) variations are computed afterwards. Variations are correctly propragated to each variable
5. input histograms that are used to compute envelopes are not removed during the hadd step, as this was causing a noticeable slowdown