Bugfix/newton linear reduction
Fix calculation of the _linearReduction when FixedLinearReduction==false. Problem appears when the initial defect is close to AbsoluteLimit. Current implementation uses MinLinearReduction instead of calculated reduction which would be higher (less strict).
In addition we solve the corner case ForceIteration==true, which accidentaly worked when the initial defect was greater than AbsoluteLimit/10, and had _linearReduction>1 otherwise. Now we use MinLinearReduction when initial defect is less than AbsoluteLimit.