Dimod

Fixes
---

- Backport docs build fix for Sphinx 4. https://github.com/dwavesystems/dimod/pull/983
- Backport Python 3.5 support removal. https://github.com/dwavesystems/dimod/pull/985
- Backport dependence on dwave-preprocessing to fix dwave-system (>1.7.0) support of both dimod 0.9 and 0.10. https://github.com/dwavesystems/dimod/pull/981

Fixes
-----
* Windows wheels now correctly include `.pxd` files. 826

New Features
------------
* `DiscreteQuadraticModel` and `SampleSet` now use the same underlying variable representation. 772, 779
* `DiscreteQuadraticModel.add_linear_equality_constraint` method. 774, 787, 798
* `AdjVectorBQM.add_linear_equality_constraint` method. 784
* Samplers now swallow unknown keyword arguments. 796, 799
* `SpinReversalTransformComposite` no longer requires a structured sampler. 802

Fixes
------
* Stopped using deprecated methods and functions to reduce the number of deprecation warnings. 773
* `DiscreteQuadraticModel.copy()` now works correctly. 777
* Fix handling of integer-valued floats in Variables. 779
* Fix the `cyDiscreteQuadraticModel.dtype` setting. 783

Fixes
-----
* `DQM.from_numpy_vectors`, `.set_quadratic` and `.set_quadratic_case` now correctly raise a `ValueError` when the user attempts to create an interaction between cases in a variable
* `BQM.from_numpy_vectors` now maintains the correct variable order
* `AdjDictBQM` biases are no longer 0 dimensional arrays when constructed from a numpy matrix
* Fixed column alignment issue when printing sample sets

New Features
------------
* C++ `AdjVectorBQM` and `AdjMapBQM` can now be constructed from COO-formatted arrays
* Python `AdjVectorBQM.from_numpy_vectors` and `AdjMapBQM.from_numpy_vectors` is now much more performant
* Cython `DiscreteQuadrataticModel` object now uses C++ COO constructors rather than an ad hoc implementation
* `gnp_random_bqm` generator for randomly connected BQMs

New Features
------------
* Documentation improvements
* Improved error messages and variable names in `DQM.from_numpy_vectors`
* `.clang-format` file for C++ formatting

Fixes
-----
* `DQM.from_numpy_vectors` no longer segfaults when given inconsistent inputs

Changes
-------
* Deprecated `SampleSet.append_variables`

New Features
------------
* `append_variables` function, replacing the funcionality of `SampleSet.append_variables`
* `append_vectors` function, to add more data vactors to a sample set
* Documentation improvements
* Support for python 3.9

Fixes
------
* Fix c++ target for windows build