InvertPSDMatrix uses dynamic matrices when using SVD

The JacobiSVD algorithm in Eigen does not accept fixed sized
matrices when performing a thin SVD. The assert enforcing this
is only triggered in non-Release builds.

So this change calls JacobiSVD with dynamically sized matrices
as a template parameter rather than a fixed size matrix.

https://github.com/ceres-solver/ceres-solver/issues/304#issuecomment-317965814

Thanks to @debalance for reporting this and @leokoppel for
providing a reproduction.

Change-Id: Ifc3d9ff20d5597f08c0f8573bf2fd99a3ed3d4d3

internal/ceres/invert_psd_matrix.h

diff --git a/internal/ceres/invert_psd_matrix.h b/internal/ceres/invert_psd_matrix.h
index 34daf88..2319fea 100644
--- a/internal/ceres/invert_psd_matrix.h
+++ b/internal/ceres/invert_psd_matrix.h
@@ -60,8 +60,7 @@
         Matrix::Identity(size, size));
   }
 
-  Eigen::JacobiSVD<typename EigenTypes<kSize, kSize>::Matrix> svd(
-      m, Eigen::ComputeThinU | Eigen::ComputeThinV);
+  Eigen::JacobiSVD<Matrix> svd(m, Eigen::ComputeThinU | Eigen::ComputeThinV);
   const double tolerance =
       std::numeric_limits<double>::epsilon() * size * svd.singularValues()(0);