Dataset interfaces

Implement class-based interfaces to datasets, both in C++ and Python domains. Replace bonzai catalogues with dataset definition files in YAML (#24 (closed)). Catalogues in the old format are also supported for backward compatibility. Drop support for baobabs.

This MR allows to specify arbitrary per-dataset settings. In principle, this means that there should be no need to rely on names of input files to adjust the logic of the processing. EWCorrectionWeight has been updated accordingly. But file names are still used for some cross-cleaning between different datasets and the classification of events based on the presence of taus.

Verified that this MR does not change histograms produced in the main analysis for a couple of files from the POWHEG gg→ZZ dataset. Checked that batch scripts created by prepareAllJobs.py were not affected as of 8de55f2e and that the latest version of prepareAllJobs.py runs successfully.

This MR needs to be integrated ahead of !25 (merged).