3.4.0

An important update with a special focus on b-tagging.

Specification of physics process in class Dataset is updated:

• A single dataset can be assigned several process codes to describe it
  at different hierarchy levels, e.g. ttbar and ttSemileptonic.
• Additional process codes have need added, some of existing ones have
  been modified. The latter change is not backward-compatible.

Support of b-tagging is greatly improved:

• Access to b-tagging efficiencies and scale factors is factorised.
  Correponding abstract base classes are introduced. New flexible file
  format to store b-tagging efficiencies.
• Reweighting for b-tagging scale factors is factorised from PECReader.
  An abstract base class to describe it is provided.
• Implementation of the adopted algorithm for b-tagging reweighting is
  improved.
• The class to perform b-tagging and classes to access b-tagging scale
  factors and efficiencies now support a symultaneous usage of all
  available working points.
• Outdated class BTagDatabase as well as corresponding example ROOT files
  with b-tagging efficiencies are removed. This change is not
  backward-compatible.

Other changes:

• Example pile-up distribution is updated to the one derived for the
  2012Bravo campaign. The old distribution is removed. This change is not
  backward-compatible.
• A new plugin FilterEventIDReminderPlugin is introduced. It filters
  events based on a reminder of division of the event number in event ID
  by a user-defined denominator. Can be useful to define the training set
  in a deterministic way.
• Several minor bugs are fixed, performance is improved.