reorganize is_fastest_rate function for NSE_NET

[zhichen3]

Based on pr#1269. I reorganized is_fastest_rate function to make it more readable and renamed it to fill_merge_indices, since it is more descriptive and accurate of what the function does.

It also gives a slight improvement in the time it takes for this function.

Here's the new profile for running Detonation on perlmutter (no mpi or anything). For whatever reason get_partition_function doesn't take that much time anymore.

Flat profile:

Each sample counts as 0.01 seconds.
  %   cumulative   self              self     total           
 time   seconds   seconds    calls   s/call   s/call  name    
 16.12     12.61    12.61 754714265     0.00     0.00  std::_Function_handler<amrex::GpuTuple<int, int> (), amrex::ReduceData<int, int>::ReduceData<amrex::ReduceOpMax, amrex::ReduceOpMax>(amrex::ReduceOps<amrex::ReduceOpMax, amrex::ReduceOpMax>&)::{lambda()#1}>::_M_manager(std::_Any_data&, std::_Any_data const&, std::_Manager_operation)
 15.43     24.68    12.07 20251079     0.00     0.00  void fcn<burn_t>(amrex::Array1D<double, 1, 2>&, amrex::Array1D<double, 1, 2>&, burn_t const&, int&) [clone .constprop.0]
  9.55     32.15     7.47 375679202     0.00     0.00  std::_Function_handler<amrex::GpuTuple<amrex::ValLocPair<double, amrex::IntVect> > (), amrex::ReduceData<amrex::ValLocPair<double, amrex::IntVect> >::ReduceData<amrex::ReduceOpMin>(amrex::ReduceOps<amrex::ReduceOpMin>&)::{lambda()#1}>::_M_manager(std::_Any_data&, std::_Any_data const&, std::_Manager_operation)
  9.08     39.25     7.10      112     0.06     0.70  Castro::react_state(amrex::MultiFab&, amrex::MultiFab&, double, double, int)
  6.66     44.46     5.21 56871959     0.00     0.00  void apply_electrons<burn_t>(burn_t&)
  4.85     48.25     3.79  5949077     0.00     0.00  Castro::okToContinue()
  4.64     51.88     3.63  5832732     0.00     0.00  in_nse(burn_t&, bool)
  3.80     54.85     2.97  5751323     0.00     0.00  void nse_hybrid_solver<burn_t>(burn_t&, double)
  3.72     57.76     2.91 14532070     0.00     0.00  void actual_eos<eos_input_t, burn_t>(eos_input_t, burn_t&)
  3.59     60.57     2.81 952020582     0.00     0.00  get_partition_function(int, tf_t const&, double&, double&)
  2.94     62.87     2.30  5762508     0.00     0.00  void hybrj<2, burn_t>(hybrj_t<2>&, burn_t const&)
  2.68     64.97     2.10  9074378     0.00     0.00  sneut5(double, double, double, double, double&, double&, double&, double&, double&)
  2.31     66.78     1.81  8736723     0.00     0.00  void evaluate_rates<0, rate_t>(burn_t const&, rate_t&)
  1.89     68.26     1.48  8736723     0.00     0.00  void fill_reaclib_rates<0, rate_t>(tf_t const&, rate_t&)
  1.12     69.13     0.88 10467305     0.00     0.00  void chabrier1998<1>(plasma_state_t const&, scrn::screen_factors_t const&, double&, double&)
  0.97     69.89     0.76  9074378     0.00     0.00  rhs_nuc(burn_t const&, amrex::Array1D<double, 1, 23>&, amrex::Array1D<double, 1, 22> const&, amrex::Array1D<double, 1, 93> const&)
  0.87     70.57     0.68 47787585     0.00     0.00  fill_merge_indices(amrex::Array1D<int, 1, 2>&, double&, int, amrex::Array1D<double, 1, 22> const&, burn_t const&, amrex::Array1D<double, 1, 93> const&, amrex::Array1D<int, 1, 22> const&, double const&)
  0.65     71.08     0.51  8736723     0.00     0.00  void rate_s32_to_p_p31_derived<0>(tf_t const&, double&, double&)
  0.64     71.58     0.50  8736723     0.00     0.00  void rate_mg24_to_he4_ne20_derived<0>(tf_t const&, double&, double&)