ThOr equivalent of SUMCONE functor

Many Run1/2 analyses are relying on isolation variables that are stored as RelatedInfo. E.g. https://gitlab.cern.ch/lhcb/Stripping/-/blob/2018-patches/Phys/IsolationTools/src/RelInfoBKstarMuMuBDT.cpp

From Mattermost discussion (@mvesteri)

(for Run3) I think this needs to be written more "functionally". Algorithm takes a signal and a container of isolation particles and produces a relations table data type with a vector of BDT inputs. BDT evaluation algorithm takes that and produces a single FP value, again as relations table data type.

Maybe first we should try to reproduce the most simple SUMCONE LoKi functor, before trying the most sophisticated BDT-equipped version

The proposed plan of actions for this issue:

• Provide a functor that calculates the sum of e.g. PT in the DeltaR cone around a candidate for particles from a given container.
• Generalise it to any functor instead of PT.
• Persist this object (create a relation Candidate->SUMCONE).
• Be able to run the MVA with SUMCONE as an input and persist it.

@tfulghes, @mvesteri, @graven, @fpolci, @rmurta