cplxpair.m: Use tolerance that depends on Z to evaluate pairing (bug #45810). * cplxpair.m: Update docstring to explain how tolerance input is calculated. Validate TOL input is a positive scalar. Use standard code to find the first singleton dimension. When determing whether a conjugate pair exists, use a tolerance of TOL*eps (abs (z(i))). Use Octave coding convestions in BIST tests. Add BIST tests for tolerance input. Add BIST tests for input validation.

%!function r = f1 ()
%!  ls = svd (1);
%!  r = eval ("ls -1;");
%!endfunction
%!function r = f2 ()
%!  [u,ls,v] = svd (1);
%!  r = eval ("ls -1;");
%!endfunction
%!function r = f3 (ls)
%!  r = eval ("ls -1;");
%!endfunction

%!test
%! ## Windows systems can't run "ls -1"
%! if (! ispc ())
%!   assert (f1 (), 0);
%!   assert (f2 (), 0);
%!   assert (ischar (f3 ()), true);
%!   assert (f3 (1), 0);
%! endif