Codex-africanus

------------------
* Convert WSClean Gaussian arcsecond and degree quantities to radians (:pr:`206`)
* Update classifiers and correct license in setup.py to BSD3 (:pr:`201`)

------------------
* Support overriding the l and m axis sign in beam_grids (:pr:`199`)
* Upgrade to python-casacore 3.3.1 (:pr:`197`)
* Upgrade to jax 0.1.68 and jaxlib 0.1.47 (:pr:`197`)
* Upgrade to scipy 1.4.0 (:pr:`197`)
* Use github workflows (:pr:`196`, :pr:`197`, :pr:`198`, :pr:`199`)
* Make CASA parallactic angles thread-safe (:pr:`195`)
* Fix spectral model documentation (:pr:`190`), to match changes in (:pr:`189`)

------------------
* Fix incorrect SPI calculation and make predict defaults MeqTree equivalent (:pr:`189`)
* Depend on pytest-flake8 >= 1.0.6 (:pr:`187`, :pr:`188`)
* MeqTrees Comparison Script Updates (:pr:`160`)
* Improve requirements handling (:pr:`187`)
* Use python-casacore wheels for travis testing, instead of kernsuite packages (:pr:`185`)

------------------
* Add a dask Estimating Progress Bar (:pr:`182`, :pr:`183`)

------------------
* Update trove to latest master commit (:pr:`178`)
* Added Cubic Spline support (:pr:`174`)
* Depend on python-casacore >= 3.2.0 (:pr:`172`)
* Drop Python 3.5 support and test Python 3.7 (:pr:`168`)
* Implement optimised WSClean predict (:pr:`166`, :pr:`167`, :pr:`177`, :pr:`179`, :pr:`180`, :pr:`181`)
* Simplify dask predict_vis code (:pr:`164`, :pr:`165`)
* Document and check weight shapes in simple gridder and degridder
(:pr:`162`, :pr:`163`)
* Restructuring calibration module (:pr:`127`)
* Upgrade to numba 0.46.0, using new inlining functionality
in the RIME and averaging code.
* Modified predict to be compatible with eidos fits headers (:pr:`158`)

------------------
* Added standalone SPI fitter (:pr:`153`)
* Fail earlier and explain duplicate averaging rows (:pr:`155`)
* CUDA Beam Implementation (:pr:`152`)
* Fix documentation package versions (:pr:`151`)
* Deprecate experimental w-stacking gridder in favour of nifty gridder (:pr:`148`)
* Expand travis build matrix (:pr:`147`)
* Drop Python 2 support (:pr:`146`, :pr:`149`, :pr:`150`)
* Support the beam in the predict example (:pr:`145`)
* Fix weight indexing in averaging (:pr:`144`)
* Support EFFECTIVE_BW and RESOLUTION in averaging (:pr:`144`)
* Optimise predict_vis jones coherency summation (:pr:`143`)
* Remove use of wraps (:pr:`141`, :pr:`142`)
* Set row chunks to nan in dask averaging code. (:pr:`139`)
* predict_vis documentation improvements (:pr:`135`, :pr:`140`)
* Upgrade to dask-ms in the examples (:pr:`134`, :pr:`138`)
* Explain how to obtain predict_vis time_index argument (:pr:`130`)
* Update RIME predict example to support Tigger LSM's and Gaussians (:pr:`129`)
* Add dask wrappers for the nifty gridder (:pr:`116`, :pr:`136`, :pr:`146`)
* Testing and requirement updates. (:pr:`124`)
* Upgraded DFT kernels to have a correlation axis and added flags
for vis_to_im. Added predict_from_fits example. (:pr:`122`)
* Fixed segfault when using `_unique_internal` on empty ndarrays (:pr:`123`)
* Removed `apply_gains`. Use `africanus.calibration.utils.correct_vis`
instead (:pr:`118`)
* Add streams parameter to dask `predict_vis` (:pr:`118`)
* Implement the beam in numba (:pr:`112`)
* Add residual_vis, correct_vis, phase_only_GN (:pr:`113`)