Major new release adding 23 new continuous colormaps from Peter Kovesi's upstream colorcet.m. The existing colormaps should all have the same names and values as in previous releases, and apart from a minor change to the type of Matplotlib categorical colormaps this release should be fully backwards compatible with previous releases.
Thanks to Randy Pittman for major improvements to the automated colormap definition code that allowed bringing in the new colormaps, to Maxime Liquet for CI and website fixes, and to Jim Bednar for docs updates, compatibility fixes, and coordinating the release.
New features:
- Added new upstream colormaps (65):
* Rainbow: `rainbow_bgyr_10_90_c83` (called `rainbow4`, a much closer match to Jet's color range than `rainbow` while remaining perceptually uniform)
* Linear: `linear_kbgoy_20_95_c57` (called `gouldian`, an [improved version of Matlab's parula colormap](https://colorcet.com/userguide/#gouldianparula)), `linear_kbgyw_10_98_c63`
* Cyclic: `circle_mgbm_67_c31`, `circle_mgbm_67_c31_s25`, `cyclic_bgrmb_35_70_c75`, `cyclic_bgrmb_35_70_c75_s25`, `cyclic_isoluminant`, `cyclic_mybm_20_100_c48`, `cyclic_mybm_20_100_c48_s25`, `cyclic_mygbm_50_90_c46`, `cyclic_mygbm_50_90_c46_s25`, `cyclic_rygcbmr_50_90_c64`, `cyclic_rygcbmr_50_90_c64_s25`, `cyclic_wrkbw_10_90_c43`, `cyclic_wrkbw_10_90_c43_s25`, `cyclic_ymcgy_60_90_c67`, `cyclic_ymcgy_60_90_c67_s25`
* Colorblind-safe: `diverging_linear_protanopic_deuteranopic_bjy_57_89_c34`, `linear_protanopic_deuteranopic_kbw_5_95_c34`, `linear_protanopic_deuteranopic_kyw_5_95_c49`, `linear_tritanopic_kcw_5_95_c22`, `linear_tritanopic_krw_5_95_c46`
- CI fixes and improvements, including testing for python 3.9 (71, 72, 75, 77)
- Clarified docs to indicate that all maps have an original list-of-numeric-triples form, typically but not always with a long name, that all those starting with `b_` are Bokeh colormaps in hex format, and that only the subset of shorter names that are aliases for longer names are Bokeh colormaps (as some other original list names like `glasbey_bw` just happen to be short (79)
Compatibility:
- For categorical (Glasbey) maps, now uses a Matplotlib ListedColormap (with distinct colors) instead of a LinearSegmentedColormap (discrete approximation to a continuous map) (79)