Mcalf

Breaking changes
* Change default Voigt implementation to `voigt_faddeeva` by ConorMacBride in https://github.com/ConorMacBride/mcalf/pull/52
To restore the old behaviour, pass `impl=mcalf.profiles.voigt.voigt_integrate` to any of the following classes or functions when you use them:
- `mcalf.models.IBIS8542Model` (after initialisation it will remember which to use for fitting and plotting)
- `mcalf.visualisation.plot_ibis8542`
- `mcalf.profiles.voigt.voigt_nobg`
- `mcalf.profiles.voigt.voigt`
- `mcalf.profiles.voigt.double_voigt_nobg`
- `mcalf.profiles.voigt.double_voigt`

New features
* Support for changing the Voigt implementation by ConorMacBride in https://github.com/ConorMacBride/mcalf/pull/51
* Publish pure Python wheels by ConorMacBride in https://github.com/ConorMacBride/mcalf/pull/53
* Upgrade supported Python versions by ConorMacBride in https://github.com/ConorMacBride/mcalf/pull/49

Documentation
* Update documentation by ConorMacBride in https://github.com/ConorMacBride/mcalf/pull/54

Testing/CI
* Revert "Fix for bug in azp template" by ConorMacBride in https://github.com/ConorMacBride/mcalf/pull/50

**Full Changelog**: https://github.com/ConorMacBride/mcalf/compare/v0.3.0...v1.0.0

New features
- Support for Python 3.9 added, with Python 3.6 dropped as well as other dependency version changes. (42)
- Stationary line core wavelength is recorded in FITS files from `mcalf.models.FitResults.save`. (47)
- MCALF version string is included in `mcalf.models.FitResults.save` files as a `VERSION` header in the primary HDU. (37)
- Add MCALF version info string at `mcalf.version.version`, using generated `_version.py`, or `setuptools_scm` for dev installs. (37)
- Added classes for managing collections of variable parameters. (38)

Refactors
- Migrate from deprecated Distutils to Setuptools. (48)

Documentation
- Include default parameter values in model signatures using features introduced in 38. (41)
- Add JOSS citation to README and documentation. (36)

Testing/CI
- Added support for skipping Azure Pipelines runs if a command such as `[skip ci]` is included in the commit message. (43)
- Remove pinned docs dependency versions. (45)
- Configured a new tox environment `codestyle` that runs `flake8` and `isort` to check the code style during an Azure Pipelines run. (46)


**Full Changelog**: https://github.com/ConorMacBride/mcalf/compare/v0.2.1...v0.3.0

New features
- Added `random_state` parameter to `mcalf.models.IBIS8542Model`, which is passed to the `sklearn.neural_network.MLPClassifier` during initialisation of the default `mcalf.models.IBIS8542Model.neural_network` attribute. This aids reproducibility. Default is `random_state=None`, so same default behaviour as previous versions of MCALF. (33)

Documentation
- Added example `Working with IBIS data` to Example Gallery, using a real IBIS spectral imaging dataset. A neural network training dataset based on this IBIS dataset is included to demonstrate the classifying capabilities. Pre-computed results for the example are also included. (31,32,34)
- Updated `LabellingTutorial.ipynb` example to run as-is with the real IBIS spectral imaging dataset. (31,32)
- Added example `Using IBIS8542Model` to Example Gallery, using randomly generated data as sample data. (27)
- Added JOSS paper to `/paper/`, and updated `.zenodo.json` to match paper title. (28,30)

New features
- Redesigned all of the plotting functions in ``mcalf.visualisation`` to be more general and integrate better with matplotlib
- Added ``mcalf.utils.plot`` module with lots of new functions
- Added function ``mcalf.utils.misc.merge_results`` to merge multiple files exported by ``mcalf.models.FitResults.save``

Refactors
- In ``mcalf.models``, refactored ``mcalf.models.ModelBase`` and ``mcalf.models.IBIS8542Model`` to make the base model more general and the IBIS model more specific
- In ``mcalf.models``, improved handling of default parameters between a model and its parent classes
- In ``mcalf.models``, moved all classes and functions from submodules into ``mcalf.models``
- In ``mcalf.models``, deprecated the inclusion of a prefilter when initialising the model in favor of processing outside of the package, however, the existing functionality remains but a warning is given
- Moved functions in ``mcalf.profiles`` to separate ``mcalf.profiles.voigt`` and ``mcalf.profiles.gaussian`` modules
- Removed unused, undocumented and untested skew normal functions from ``mcalf.profiles``
- Moved functions in ``mcalf.utils`` to separate submodules
- Merged all submodules of ``mcalf.visualisation``

Documentation
- Docstrings are made more consistent, typos are corrected and more examples are added
- Integrated ``sphinx-gallery`` package into documentation
- Added examples to the example gallery for all functions in ``mcalf.visualisation``
- Added detailed documentation on running tests on the code
- Added citation for new paper ([doi:10.1098/rsta.2020.0171](https://doi.org/10.1098/rsta.2020.0171))
- In ``mcalf.models``, generated docstrings from replacement strings to improve consistency
- In ``mcalf.profiles.voigt``, generated docstrings from replacement strings to improve consistency
- Added intersphinx mappings to other packages' documentation
- Included private members of ``mcalf.models.ModelBase`` in subclasses (https://github.com/astropy/sphinx-automodapi/pull/121)
- Automatically update the year in documentation

Testing/CI
- Increased test coverage throughout the package
- Using ``pytest-mpl`` to verify matplotlib figures produced during tests
- Migrated testing from Travis CI to Azure Pipelines
- Moved tests from the package root to ``mcalf.tests`` inside the package source directory to enable testing wheels
- Moved details from ``setup.py`` to ``setup.cfg``
- Moved testing and documentation dependencies to separate extra requirements sections
- Added custom configuration settings for pytest, coverage and tox
- Using ``setuptools_scm`` to keep track of package version number and files
- Force ``mcalf.models.voigt`` to not use the C library if running on readthedocs as they do not support building C extensions

- Added documentation
- Added tests and code coverage checking
- Added support for Windows
- Added ability to export results to FITS files
- Various minor bug fixes and code improvements
- Migrated from deploying to PyPI using Travis CI to Azure Pipelines such that wheels can be more easily built
- MCALF will be available on conda-forge

This is the first release of MCALF