DGP parameters instead of actual best fit as solution

[LouisSirugue]

Issue:

In reg_simple, reg_simple_arrow, and other fitting apps, the squared errors/the arrows turn green when we tune the sliders to the parameters from the DGP, not from the regression on the randomly drawn sample.

Example:

In reg_simple, the parameters of the DGP a = -2 and b = 1.5 make the squared errors turn green, but we get a lower sum of squared errors with a = -1.4 and b = 1.3, as they fit better the randomly drawn sample.