Change SVD algorithm in covariance computation.

Switch from JacobiSVD to BDCSVD in
ComputeCovarianceValuesUsingDenseSVD. This should increase
the performance for larger covariance matrices. See
https://eigen.tuxfamily.org/dox/classEigen_1_1BDCSVD.html

Change-Id: Icde4dec89f506b638b0f9f1aee3b7cfc9e4d72fc

internal/ceres/covariance_impl.cc

diff --git a/internal/ceres/covariance_impl.cc b/internal/ceres/covariance_impl.cc
index 299dda8..6c26412 100644
--- a/internal/ceres/covariance_impl.cc
+++ b/internal/ceres/covariance_impl.cc
@@ -752,8 +752,8 @@
   }
   event_logger.AddEvent("ConvertToDenseMatrix");
 
-  Eigen::JacobiSVD<Matrix> svd(dense_jacobian,
-                               Eigen::ComputeThinU | Eigen::ComputeThinV);
+  Eigen::BDCSVD<Matrix> svd(dense_jacobian,
+                            Eigen::ComputeThinU | Eigen::ComputeThinV);
 
   event_logger.AddEvent("SingularValueDecomposition");