The post-linear fit should use the linear fit parameters as starting …

…values.

This makes sure that when using the photodiode (much larger scaling) that
things converge correctly when checking residuals.

python/lsst/cp/pipe/cpLinearitySolve.py

@@ -685,7 +685,7 @@ def run(self, inputPtc, dummy, camera, inputDims,
                 residuals = np.full_like(linearizeModel, np.nan)
             else:
                 postLinearFit, _, _, _ = irlsFit(
-                    [0.0, 100.0],
+                    linearFit,
                     inputAbscissa[mask],
                     linearizeModel[mask],
                     funcPolynomial,