Commit svdx files sent by Osni on Oct 31st 2015

Commit  svdx files sent by Osni on Oct 31st 2015
Updated Makefile
Updated file format (Oxygen)
Small Fix: sdrvbd / cdrvbd has a lot of double / double complex in comments

From Osni:
Files for LAPACK/SRC (all new, additions to Makefile are needed):

 - dbdsvdx.f, sbdsvdx.f: full or partial (subset) SVD of a bidiagonal matrix through an associated eigenvalue problem
 - dgesvdx.f, sgesvdx.f, zgesvdx.f, cgesvdx.f: full or partial (subset) SVD of a general matrix by invoking bdsvdx

Files for LAPACK/TESTING/EIG :
- dchkbd.f, schkbd.f: added tests 20-34 for bdsvdx
- dbdt04.f, sbdt04.f (new): needed for tests 25 and 30 in chkbd
- dlahd2.f, slahd2.f: added information about tests 20-34 in chkbd
- derrbd.f, serrbd.f: added tests for the values of INFO returned by bdsvdx
- ddrvbd.f, sdrvbd.f, zdrvbd.f, cdrvbd.f: added tests 23-35 (real case) and 15-27 (complex case) for gesvdx.
- dbdt05.f, sbdt05.f, zbdt05.f, cbdt05.f (new): needed for tests 30,33 (real case) and 22,25 (complex case) for gesvdx
- derred.f, serred.f, zerred.f, cerred.f: added tests for the values of INFO returned by gesvdx

Current Issues:
 - 16 real tests do not pass the threshold ( ssvd.out) - but seems to go through in debug
 - serrbd.f has some DOUBLE PRECISION calculation inside - Is that ok or shall we change it to REAL?
Line 113-119
*     Set the variables to innocuous values.
*
      DO 20 J = 1, NMAX
         DO 10 I = 1, NMAX
            A( I, J ) = 1.D0 / DBLE( I+J )
   10    CONTINUE
   20 CONTINUE

TODO:
     - corresponding LAPACKE routines
     - Test Doxygen format