Neml2

Increments version number and (hopefully) releases to PyPI.

What's Changed
* Python bindings now available. This release will be made avialable on PyPi.
* Crystal plasticity models added and verified against NEML.
* Generic nonlinear parameters, i.e. for temperature dependent material models.
* Gurson-Tvergaard-Needleman model
* Various improvements to the build system
* Added a polynomial ROM example

**Full Changelog**: https://github.com/reverendbedford/neml2/compare/v1.2.0...v1.4.0

This release contains various bug fixes and feature improvements. Most notably, this release is accompanied by an ANL report authorized for release. The report was automatically generated using doxygen with tag v1.2.0. It serves as a static reference manual and also a starting point for technical documentation -- future ANL reports on NEML2 will mostly focus on changes and improvements and will not repeat the same set of notes over again.

Other notable changes include
- The `Predictor` system is removed. Instead, we let the "driver" programs handle predictor calculation. An example predictor implementation is available in `TransientDriver`.
- An initial effort in integrating NEML2 into [BlackBear](https://github.com/idaholab/blackbear), an application based on the finite element method. A significant speed-up can be achieved by offsetting constitutive updates onto GPU(s) using NEML2. The integration is still being worked on, and is expected to complete in the next few releases.

This release accomplishes most of the basic goals set out for the NEML2 library:
- Easy ability to construct a model and send it to a torch device (GPU) for execution.
- Input file format using the [hit](https://mooseframework.inl.gov/source/parser/Parser.html) parser
- Automatic construction of (static) function trees for models, automatically caching intermediate results and constructing the overall model Jacobians from individual model partial derivatives
- Allows the user to freely choose between AD and hard coded implementations of the model partial derivatives
- An interface to get the parameter partials of complete models for optimization

Future releases will concentrate on:
- Integrating NEML2 into constitutive model training frameworks, like [pyoptmat](https://github.com/Argonne-National-Laboratory/pyoptmat)
- Adding "heavier" constitutive models to demonstrate the benefits of vectorizing the constitutive model update
- Additional examples of applying NEML2 to multiphysics problems
- Better integration into the [MOOSE](https://mooseframework.inl.gov/) framework

Initial public release of NEML2. This includes basic functionality and some common solid mechanics constitutive models as tests.