Raysect

---------------------------

API changes:
* Both extract_rotation() and extract_translation() now return ctuples in cython.
* The extract_rotation() string z_axis parameter has been replaced with a boolean z_up parameter for switching coordinate systems.

Bug fixes:
* extract_rotation() now returns the correct values for yaw, pitch and roll in the z axis up coordinate system (Jamie Bayne).
* The quiet parameter is now correctly handled by Observer1D and Observer2D.
* Fixed Ray.extinction_min_depth bug that prevented the attribute being set to 1.
* Fixed pipeline figure title rendering due to matplotlib canvas.set_window_title() deprecation.
* Fixed __setstate__() bug in SpectralPowerPipeline that prevented unpickling.

New:
* Added cython pxd files for all observers.

---------------------------

API changes:
* Intersection object is now passed to evaluate_surface, user code will need to be updated to include the additional parameter.

Bug fixes:
* Spectrum samples attribute is now readonly to prevent users replacing the array and potentially causing a segfault.

New:
* Added pyproject.toml (Jack Lovell)
* Mesh returns a specialised MeshIntersection object that exposes the triangle and barycentric coordinates of the intersection.

--------------------------

Bug fixes:
* MulticoreEngine fixes for MacOS and python 3.8+
* The spectral pipeline 'display_progress' attribute was not accessible from python.
* The pipeline line 'name' attribute was not accessible from python.
* Fixed half-pixel offset in multiple cameras due to an index offset error.
* Improved normalisation in quaternion_to().

New:
* Added cython factory functions for quaternions.
* Tetrahedral mesh interpolation function.
* 1D, 2D and 3D array interpolation functions.
* 1D, 2D and 3D blend functions.

--------------------------

Bug fixes:
* Fixed the last integration segment not contributing to the integral in NumericalIntegrator.
* Incorrect construction of lens primitives when lens long barrels are requested.

API Changes:
* The core.math.function package has been redesigned to introduce functions with different return types. The existing FunctionXD classes now live under core.math.function.float.

New:
* Added core.math.function.vector3d.FunctionXD objects which return Vector3D objects.
* Added Quaternions and associated transform utility functions.
* Added utility functions for extracting the rotation and translation from an AffineMatrix3D.
* A new Cornell Box demo with a virtual camera with a Cooke triplet lens assembly and a simulated CCD.

--------------------------

New:
* Added relational operators to the FunctionXD objects.
* Added abs() operator to FunctionXD objects.
* Square root (SqrtXD) and error functions (ErfXD) added to FunctionXD package.
* Added SpectralAdaptiveSampler1D and SpectralAdaptiveSampler2D.
* FullFrameSampler2D, RGBAdaptiveSampler2D, MonoAdaptiveSampler2D now supports pixel masks directly.

Bug fixes:
* MulticoreEngine render engine now passes back exceptions from worker processes and correctly terminates the render.

---------------------------

API Changes:
* Blend material modifier now blends the surface and volume by default. Use the surface_only and volume_only options to configure.
* Cython minimum, maximum and peak to peak methods can now cope with arbitrary strides.
* Interpolator2DMesh and Discrete2DMesh have import has changed from raysect.core.math.interpolators to raysect.core.math.function.

New:
* Added an Add material modifier than sums the contributions from two materials.
* Added emissivity function to BlackBody.
* Extended Spectrum cython mathematical methods to simplify combining Spectrum objects.
* Expanded the function framework:
- Constants are now autowrapped as ConstantXD objects automatically.
- Added PowXD, ExpXD, SinXD, CosXD, TanXD, AsinXD, AcosXD, AtanXD, Atan4QXD functions.
- Argument pass-through supported with ArgXD functions.
- Interpolation functions moved under raysect.core.math.function.
* Improved the Multicore render engine by batching work to reduce IPC overhead. Renders with smaller amounts of work per pixel are now substantially quicker.
* Improved multiple demos and related documentation.