bugfix in thirdDerivativeOfDistanceSquaredWRTFirst1AndSecond2Argument.  Now it produces the same result as the fd approximation at least in for S^1

[[Imported from SVN: r7066]]