Add a history scheme to the DatasetCollection to track dataset revisions in a more standardized way

[martwo]

Currently the dataset revisions (usually patch numbers) are just tracked by their numbers, but what each patch number means in terms of actual change to the dataset is not always described in the dataset definition itself. Usually only in the README file of the dataset on disk.
I propose a history scheme to keep track of these revisions in the dataset definition itself. So that the user could do
print(DatasetCollection.history)
2020-04-05: p01 -> p02:
Add the additional observables: TUM_hybrid, TUM_sigmaBDT
2020-04-02: p00 -> p01:
Remove energy cut on TruncatedEnergy at 200 GeV

It should be noted that the key of the history entry should be the date and not the patch number, because in general something else than the patch number could have changed.