======================================================================

   THIS CONTRIB IS DEPRECATED. PLEASE USE THE SoftDrop CLASS IN
                     RecursiveTools INSTEAD

======================================================================



Contrib implementing the SoftDrop declustering introduced by Andrew
Larkoski, Simone Marzani, Gregory Soyez, and Jesse Thaler in
arXiv:YYMM.NNNN.

The method proceeds as follows:

 0. recluster the jet with the Cambridge/Aachen algorithm

 1. undo the last step of the clustering step j -> j1 + j2.
 
 2. if the two following conditions are satisfied 
      (i) there is a soft frop, i.e. \f${\rm
          min}(p_{tj1},p_{tj2})/(p_{tj1},p_{tj2}) > z_{/rm cut}
          (\Delta R_{j1,j2}/R_0)^\beta\f$,
     (ii) there is a mass drop, i.e. \f$m_{j1}/m_{j} < \mu\f$
    keep j as the result of the tagger (with j1 and j2 its 2
    subjets)

    In the above conditions, zcut, beta and mu are parameters of
    the tagger and R0 is the radius parameter of the original jet.

 3. otherwise, redefine j to be equal to the largest-mt of j1 and
    j2 and return to step 1.

----------------------------------------------------------------------
Options
----------------------------------------------------------------------

The constructor has the following arguments:
  - The first  argument is the angular exponent beta 
  - The second argument is the soft-drop cut (zcut) [0.1 by default]
  - The third  argument is the minimal mass drop that is required
     (mu) [1 by default]

Note that the arguments are in an opposite order as the
MassDropTagger. This allows to keep mu=1 as a default.

----------------------------------------------------------------------
Input conditions
----------------------------------------------------------------------

 - the original jet must have constituents

 - the current version requires the jet to have an associated
   ClusterSequence.

   Note that for future versions we intend to also support more
   generic jets [e.g. CompositeJets]. [Note for developpers: in
   practice, this means being a bit more careful with the
   reculstering + allowing the user to specify R0 explicitly]

----------------------------------------------------------------------
Output/structure
----------------------------------------------------------------------

 - the 2 subjets are kept as pieces if some substructure is found,
   otherwise a single 0-momentum piece is returned
 - the 'mu' and 'z/(\Delta R_{j1,j2}/R_0)^\beta' values
   corresponding to the unclustering step that passed the tagger's
   cuts

----------------------------------------------------------------------
Open question(s) [for developpers]: 
----------------------------------------------------------------------
 - Q: R0 can be undefined in some cases [e.g. if the input jet is a
      composite jet]?
   A: In a future version, support more generic inputs. For these
      cases, add a 2nd ctor where one specifies R0.

 - Q: do we have a set_subtractor for interaction with PU subtraction?
   A: Yes [to be implemented!]