Refnx

------
- Add a little machinery for energy dispersive calculations. This is
implemented by adding a wavelength attribute to Structure. The
Structure.slabs method then works out the Slab representation for that
particular wavelength. The wavelength attribute should be used by the
Component.slabs method when asking for the SLD of a Scatterer. The Slab and
MixedSlab Components have been altered to permit this. User provided
components should work as they currently do, but probably have to be
rewritten to permit energy dispersive operation.
For energy dispersive operation classes that inherit Scatterer should
override the newly introduced Scatterer.complex() method.

------
- change wheel builder to make 3.10 wheels.
- check that pyinstaller based macOS standalone app runs in CI.
- update install documentation
- workaround for difficulty using orsopy on Python 3.10
- added fields to ReflectReduce output dict, including omega, reflected_beam,
direct_beam.

------
- Add NPY_NO_DEPRECATED_API=NPY_1_7_API_VERSION define to cython extension
building to remove warnings. This requires a different mode of access to
the array data.
- Fixed drag and drop of datasets onto refnx GUI.
- Add OrsoDataset and a generalised load_data function.
- refnx GUI removes horizontal scrollbar.
- Fixed crash when trying to access version numbers when saving expt file.

------
- use oldest-supported-numpy to ensure that older numpy versions are used to
build refnx wheels. This will allow refnx to be used with more recent
versions as well as the oldest version supported on a given Python version.
Previously refnx wheels were built with the newest numpy available at
build time, and were probably not compatible with older numpy versions.
- Rejig the build process to make macOS universal wheels.

------
- bugfix: remove printing of auxiliary parameters in Objective.setp. A remnant
print function was in that method. However, those parameters aren't present
in a global objective, which caused an AttributeError.

------
- Added polarisation reduction (Oliver Paull)
- Introduce Objective.auxiliary_params. These are extra parameters that are
modified during the calculation of Objective statistics, and are varied
during a fit, but may not be directly part of Model calculating a signal.
Their main purpose is to aid in creating constraints.
- Add example to show how to co-refine non-spinflip NR data.
- Silenced warning during test from slow reflectivity calculation, which
erroneously gave the impression that the C based kernel wasn't available.