the various LAB spaces are supposed to be *perceptually linear* which is computationally suprisingly nonlinear (thats also the reasons why there are so many different models, which basically try to be more linear than the others).

Writing a node for converting color values to e.g. CIE 1976 L*a*b* would be completely possible and cool and would make an excellent learning exercise.

But you were actually asking for a color transformation node - color transformations are internally using a 4x4 matrix multiplication. this matrix multiplication is able to do a suprisingly useful number of basic color manipulation things - but you can not express a conversion from RGB to a lab space with it. Our brain is way to nonlinear for this.