automatic centering / rescaling of covariates (and data?)

[lisa-gm]

Increase numeric stability by rescaling the covariates (i.e. projection matrix A (column-wise) internally before starting the computations and then rescaling back before returning any results. I haven't thought through all of the details.

[vincent-maillou]

Isn't this issue related to: Unified PyINLA/Core Compute Scale of Hyperparameters and Model Scale of the Hyperparameters #6 ?

Is it related to #130 in some way or is it a different scaling than the "user" vs "compute" scales?