Jsonpickle

======
* The ``ignorereserved`` parameter to the private ``_restore_from_dict()``
function has been restored for backwards compatibility. (+501)

======
* Nested dictionaries in `py/state` are now correctly restored when
tracking object references. (+501) (500)

======
* `jsonpickle.ext.numpy.register_handlers` now provides options that are forwarded
to the `NumpyNDArrayHandler` constructor. (+489)
* Fix bug of not handling ``classes`` argument to `jsonpickle.decode`
being a dict. Previously, the keys were ignored and only values were
used. (+494)
* Allow the ``classes`` argument to `jsonpickle.pickle` to have class
objects as keys. This extends the current functionality of only having
class name strings as keys. (+494)
* The ``garden setup/dev`` action and ``requirements-dev.txt`` requirements file
now include test dependencies for use during development.
* Added support for Python 3.13. (+505) (504)

======
* Fixed an issue with django.SafeString and other classes inheriting from
str having read-only attribute errors (478) (+481)
* The test suite was made compatible with `pytest-ruff>=0.3.0`. (+482)
* A `garden.yaml` file was added for use with the
`garden <https://crates.io/crates/garden-tools>_` command runner. (+486)
* The test suite was updated to avoid deprecated SQLALchemy APIs.
* The `jaraco.packaging.sphinx` documentation dependency was removed.

======
* Compatibilty with Pandas and Cython 3.0 was added. (460) (+477)
* Fixed a bug where pickling some built-in classes (e.g. zoneinfo)
could return a ``None`` module. (447)
* Fixed a bug where unpickling a missing class would return a different object
instead of ``None``. (+471)
* Fixed the handling of missing classes when setting ``on_missing`` to ``warn``
or ``error``. (+471)
* The test suite was made compatible with Python 3.12.
* The tox configuration was updated to generate code coverage reports.
* The suite now uses ``ruff`` to validate python code.
* The documentation can now be built offline when ``rst.linker`` and
``jaraco.packaging.sphinx`` are not available.

======
* Properly raise warning if a custom pickling handler returns None. (433)
* Fix issue with serialization of certain sklearn objects breaking when
the numpy handler was enabled. (431) (+434)
* Allow custom backends to not implement _encoder_options (436) (+446)
* Implement compatibility with pandas 2 (+446)
* Fix encoding/decoding of dictionary subclasses with referencing (+455)
* Fix depth tracking for list/dict referencing (+456)