This reverts commit 1d26ec6.

this still only works with aprox19

…tro#1282)

these include Make.Microphysics_extern
but not Make.Microphysics

Here I separated get_nse_state into two separate functions, one that calculates the exponent part of the nse mass fraction (depends on chemical potentials) and the other calculates components that do not depend on chemical potentials.

This way we don't have to compute unnecessary parts when we solve for chemical potentials, mu_p and mu_n during hybrid solver, which includes getting the partition function.

Heres the new profile, it brings down # of calls for get_partition_functions from 900,000,000+ (from pr1274) to 200,000,000+

abort earlier if T_in < T_nse_net to avoid unnecessary computing.

This updates ECSN with the latest pynucastro, which changes the rate tables to be log10.  Due to differences in the units conversion factors and the number of decimals between the existing tabular rates and the ones of the pynucastro PR AMReX-Astro#550, the existing benchmark will differ.

The goal of this change is to remove the global data from actual_network_data.cpp, which in turn allows elimination of that file. Rather than an array, this can be handled with a constexpr function emulating the array. The new function is templated to allow compile-time evaluation of the binding energy. However, not all locations in Microphysics will be readily available to switch to a compile-time evaluation, so a "legacy" interface which accepts the species index at runtime is also provided. (This requires constexpr_for to be available in the network, so it is moved to a separate utility location.)