SVD of a matrix based on Lapack interface

SVD_LAPACK computes the singular value decomposition of a matrix by calling LAPACK subroutines.

SVD_LAPACK provides a SVD routine based on ZGESDD,
which is not supported as Matlab's built-in function, at least, explicitly.
Lapack, a fortran computational library, has two different subroutines
for the Singular Value Decompostion (SVD): xGESVD and xGESDD.
xGESVD is based on an implicit QR iteration and xGESDD uses a
divide-and-conquer approach.
See <http://www.netlib.org/lapack/lug/node32.html> and
<http://www.netlib.org/lapack/lug/node53.html> for Lapack subroutines.
Matlab's built-in function svd seems to use the lapack subroutine xGESVD.
Meanwhile, Octave's built-in function svd support both algorithms
by using svd_driver().

If people want to use xGESDD routines in Matlab,
it is required to write a MEX-file to call the routine.
See <http://www.mathworks.co.kr/matlabcentral/newsreader/view_thread/250196>
But, it is not easy to call the lapack properly, because it needs a good
knowledge of fortran work space.

In this submission, SVD_LAPACK function provides both routines and the lapack interfaces.
It could be an example for calling SVD rountine from Lapack interface.
Although it supports double precision routines, they can be replaced
by single precision routines without any problem.