Requesting an Oklab color space
CIE Lab is known to have major issues in the blue hue department (see gimp#3580 (closed)). A new proposed color space called Oklab fixes this issue and fits CIECAM16 data better.
The matrix-nonlinearity-matrix structure is identical to that of CIE Lab, although the nonlinearity is simpler being just a cube. It is intended to be used anywhere CIE Lab was used (standard dynamic range), but with improved hue and lightness prediction.
When implemented using floats, Oklab is naturally expected to be faster than CIE Lab with an LUT correction.
See also https://raphlinus.github.io/color/2021/01/18/oklab-critique.html.
I went for trying to implement the thing, and the main issue seems to be that babl_space_{from,to}_xyz{,f}
gives a D50-adapted value, while Oklab expects a D65-adapted one. I should modify the matrix to multiply the input by another matrix, but babl_chromatic_adaptation_matrix
is private.