Simplify class handling ('single' vs. 'double') for statistical distribution random functions. * betarnd.m, cauchy_rnd.m, chi2rnd.m, exprnd.m, frnd.m, gamrnd.m, geornd.m, lognrnd.m, nbinrnd.m, normrnd.m, poissrnd.m, trnd.m, unidrnd.m, unifrnd.m, wblrnd.m: Use class argument to random number function rather than explicitly or implicitly casting output to class single.

      subroutine xdnrm2 (n, x, incx, retval)
      double precision dnrm2, x(*), retval
      integer n, incx
      retval = dnrm2 (n, x, incx)
      return
      end