Pynucastro

* fix the zenodo authors (898) and automate the zenodo bibtex in
the docs (880)

* codespell fixes (897)

* AmrexAstroCxxNetwork now ifdefs out neutrinos (884)

* officially support python 3.13 (860)

* update package requirements (896)

* Implement Debye-Huckel screening (870)

* Fix a CODATA constants inconsistency between Nubase and SciPy
(877) and compatibility with SciPy < 1.15.0 (889)

* Address a new warning in matplotlib 3.10.0 (879, 885)

* Bump up jinja2 version (871)

* Added the ability to approximate A(n,gamma)X(n,gamma)B into an
effective rate A(nn,gamma)B (818)

* `SimpleCxxNetwork` now includes electron screening using the
prescription in Chugunov 2007 (844)

* added "NSE protons" that evaluate the same as a proton but can be
used to decouple light and heavy reactions (853, 854, 862, 863)

* `modify_products` was moved to the base `Rate` class all rate classes
can use it (851)

* Fix an issue with AMD HIP/ROCm memory access on the GPU in the
partition functions (859)

* A `FortranNetwork` is now available that provides a wrapper around
`SimpleCxxNetwork` with Fortran interfaces. (841, 847)

* `Nucleus` objects can now be added and subtracted (835)

* Fix a compiler issue with `SimpleCxxNetwork` and ReacLib weak rates
(846)

* Documentation improvements (782, 831, 832, 843, 848, 852,
855, 858, 861, 864, 866, 867, 874, 881, 887, 888),
including new examples (783, 800, 826) and build enhancements
(893)

* Clean-up unused class data in `SimpleCxxNetwork` (845)

* Clean up the rate module by separating it into submodules (840, 842)

* Add the ability to label edges in network plots (837)

* Clean-up the `make_ap_pg_approx` code (830, 850)

* `PartitionFunction.eval` now returns a python float instead of a
numpy float (839)

* Rates now store their source from publications (822, 834)

* Change some methods to properties in `Composition`, `Library`, and
`Rate` (829)

* `PythonNetwork` rates now can take rho and Y. (828)

* CI fixes and updates (827, 890)

* Child rates in `ApproximateRate` are now stored as a dictionary
(824) + additional cleaning (812)

* Allow the identical particle factor in `Rate` to be disabled (821)

* `Rate.eval()` now take rho and comp (820)

* Update Zenodo authors (815)

* Use python's `pathlib.Path` throughout (811)

* `Nucleus` now stores the halflife (801)

* expanded documentation on exporting networks (805) and physical
constants (793)

* some doc test fixes (806, 807, 809)

* expanded unit testing (802)

* update pylint disabled features (794)

* dropped support for python 3.9 (787)

* we now compute the Q value for rates from masses instead of
binding energies if it is not available (797)

* updated the nuclear properties to all be based on Nubase 2020,
with the mass excesses being the primary quantity used and binding
energies now derived from the mass excesses. (788, 792, 796)

* added a `get_all_nuclei()` method that returns every nucleus for
which we have a mass (798)

* removed unused scripts in the nucdata module (789)

* fixed the labels for trace nuclei in a `Composition` pie chart
(781)

* added a warning if we use the ReacLib n decay rate (778)

* doc updates, including an example of dealing with duplicate rates
(777), new theme (775), axis label fixes (774), some examples
of nuclear astrophysics concepts (762, 766)

* fix the limits for `TabularRate` plots (779)

* created a new `alternate_rates` module and added the de Boer et
al. 2017 C12(a,g)O16 rate (769)

* the minimum version of python supported is now 3.8 (768)

* some test fixes (755, 757, 760, 763)

* C++ autodiff changes / cleanups (752, 759)

* `AmrexAstroCxxNetwork` now create a ydot_weak function that
evaluates the weak rate contributions to compute dYe/dt (739)

* numpy 2.0 support added (748)

* new logo! (744, 746)

* pynucastro can now be installed by conda (741, 743)

* minor test fixes (729, 735, 740, 745)

* doc fixes (737)

* pylint fixes (733)

* Composition now allows for direct key-value indexing (731),
and has a new `set_random()` method (728)

* The `screen5` screening routine has been added (730)

* C++ partition function interpolation is now cached (726)

* Switch to using `math.fsum()` instead of `sum()` for better python
3.12 compatibility (720)

* Some modernization of the AMReX-Astro C++ code output (709, 714,
717)

* Documentation improvements, including references and contributing
instructions, and NSE table (669, 679, 695, 697, 706, 707)

* Clean-up up some plot interfaces to always take rho, T, comp in
that order (693)

* Performance improvements (680, 685, 692)

* A new `constants` module was added to provide fundamental constants
(688)

* A new `NumpyNetwork` was added for use with the rate reduction
algorithm (684)

* Neutrino cooling routines based on Itoh et al. 1996 were added
(628)

* A new method for generating an NSE table was added (676)

* A new `NSENetwork` class was added for dealing with NSE (675)

* Moved some screening utilities out of `RateCollection` (666)

* A `PythonNetwork` now write out a function to compute the energy
release (640)

* `RateCollection.evaluate_energy_generation()` can now return the
neutrino losses from weak rates separately. (665)

* `AmrexAstroCxxNetwork` now capitalizes the species names to be
consistent with other AMReX-Astro Microphysics networks (652)

* The pynucastro project configuration was updated and reorganized
to `pyproject.toml` (657, 660, 661)

* `Library.get_rate_by_name` no longer returns None if one of
the input rate names is not found (659)

* A C++ indexing error at the boundary of a weak rate table was
fixed. (654)

* The C++ `AmrexAstroCxxNetwork` for evaluating weak rate neutrino
losses was cleaned up to fix a bounds error (655)

* We now explicitly check in the C++ code if a weak rate table
if found and contains data (654)

* a `RateCollection` now enforces that there are no duplicate rates. A
`find_duplicate_links()` method has been added to `Library` to allow
one to construct a `Library` without duplicates and then create a
`RateCollection` or other network from that. (651)

* Some documentation updates (643, 650)

* The positron rates in the Langanke weak rate tables have been
merged with the corresponding electron rates and the n <-> p rate
was added (648)

* The `Composition.bin_as()` method now takes a list of nuclei
to exclude from being binned into (642)

* `RateCollection.evaluate_ydots()` now can take a filter to
only evaluate a subset of the rates. (641)