Make scalingfactors compatible with linearisation

[simulutions]

since #2508 , testitem "Allocation example model" is marked as broken.

The problem is that in this test model, the linearisation combined with the scaling factors result in matrix coefficients of 5.33e-11. This is probably numerically instable and highs marks the problem as infeasible. If we set the scaling factors to 1, the problem is actually better scaled and solves with no problem. However, if we do this other unit tests fail. So we need a solution such that all problems are scaled properly