Remove unnecessary (and possibly harmful code to set FAILURE flag to …

…False

I think this is leftover from an older style of setting flag, which allowed
the algorithm to throw out from anywhere, knowing that the flag was True.
I believe that such failures will now be caught by the framework.
Plus, this strategy would circumvent any other code (such as the Centroid
Checker itself) which might want to set a failure flag but allow the alg.
to continue on.

src/NaiveCentroid.cc

@@ -111,7 +111,6 @@ void NaiveCentroidAlgorithm::measure(
     result.x = lsst::afw::image::indexToPosition(x + image.getX0()) + sum_x / sum;
     result.y = lsst::afw::image::indexToPosition(y + image.getY0()) + sum_y / sum;
     measRecord.set(_centroidKey, result);
-    _flagHandler.setValue(measRecord, FAILURE, false);  // if we had a suspect flag, we'd set that instead
     _centroidChecker(measRecord);
 }
 

src/SdssCentroid.cc

@@ -536,7 +536,6 @@ void SdssCentroidAlgorithm::measure(
     result.xSigma = sqrt(dxc*dxc);
     result.ySigma = sqrt(dyc*dyc);
     measRecord.set(_centroidKey, result);
-    _flagHandler.setValue(measRecord, FAILURE, false);
     _centroidChecker(measRecord);
 }