Make the parallel reduction when applying the Wells.

The B matrix is basically a component-wise multiplication
with a vector followed by a parallel reduction. We do that
reduction to all ranks computing for the well to save the
broadcast when applying C^T.

opm/simulators/wells/WellInterface_impl.hpp

@@ -30,6 +30,7 @@ namespace Opm
     template<typename TypeTag>
     WellInterface<TypeTag>::
     WellInterface(const Well& well,
+                  const ParallelWellInfo& pw_info,
                   const int time_step,
                   const ModelParameters& param,
                   const RateConverterType& rate_converter,
@@ -40,6 +41,7 @@ namespace Opm
                   const int first_perf_index,
                   const std::vector<PerforationData>& perf_data)
       : well_ecl_(well)
+      , parallel_well_info_(&pw_info)
       , current_step_(time_step)
       , param_(param)
       , rateConverter_(rate_converter)