Fix(xc): disable SF builtin dispatch for gga_grad=3

SF builtin (xc_functional_gga_noncol_sf_builtin.cpp) produces
non-convergent results with varying ecutwfc due to numerical
instability from independent FFT sequence divergence.

  gga_grad=3 - gga_grad=1:
    ecut=20:  -0.071 eV
    ecut=35:  -0.003 eV
    ecut=50:  +0.291 eV  (sign flip!)

Remove dispatch from v_xc so gga_grad=3 falls through to the
numerically stable gradcorr path. SF builtin implementation
retained for future debugging.

Also fix gcc_spin guard and zeta clipping in SF builtin to
match gradcorr conventions (prep for future re-enablement).

Author: dyzheng

source/source_hamilt/module_xc/xc_functional_gga_noncol_sf_builtin.cpp

@@ -135,13 +135,8 @@ std::tuple<double, double, ModuleBase::matrix> v_xc_ncgga_sf_builtin(
             {
                 double zeta = (rhotmp1[ir] - rhotmp2[ir]) / rh;
                 zeta = std::fabs(zeta);
-                if (zeta > 1.0 - epsr)
-                    zeta = 1.0 - epsr;
                 const double grh2 = (gdr1[ir] + gdr2[ir]) * (gdr1[ir] + gdr2[ir]);
-                if (std::sqrt(std::abs(grh2)) > 1e-10)
-                {
-                    XC_Functional::gcc_spin(rh, zeta, grh2, sc, v1cup, v1cdw, v2c);
-                }
+                XC_Functional::gcc_spin(rh, zeta, grh2, sc, v1cup, v1cdw, v2c);
             }
 
             vup_gga[ir] = e2 * (v1xup + v1cup);
@@ -287,13 +282,8 @@ void gradcorr_ncgga_sf_builtin(const Charge* const chr, ModulePW::PW_Basis* rhop
         {
             double zeta = (rhotmp1[ir] - rhotmp2[ir]) / rh;
             zeta = std::fabs(zeta);
-            if (zeta > 1.0 - epsr)
-                zeta = 1.0 - epsr;
             const double grh2 = (gdr1[ir] + gdr2[ir]) * (gdr1[ir] + gdr2[ir]);
-            if (std::sqrt(std::abs(grh2)) > small)
-            {
-                XC_Functional::gcc_spin(rh, zeta, grh2, sc, v1cup, v1cdw, v2c);
-            }
+            XC_Functional::gcc_spin(rh, zeta, grh2, sc, v1cup, v1cdw, v2c);
             v2cup = v2c;
             v2cdw = v2c;
             v2cud = v2c;

source/source_hamilt/module_xc/xc_functional_vxc.cpp

@@ -22,15 +22,6 @@ std::tuple<double, double, ModuleBase::matrix> XC_Functional::v_xc(const int& nr
 {
     ModuleBase::TITLE("XC_Functional", "v_xc");
 
-    if (PARAM.inp.nspin == 4 && (PARAM.globalv.domag || PARAM.globalv.domag_z) && PARAM.inp.gga_grad == 3)
-    {
-        const int ft = XC_Functional::get_func_type();
-        if (ft == 2 || ft == 4)
-        {
-            return ModuleXC::NCGGA_SF_Builtin::v_xc_ncgga_sf_builtin(nrxx, ucell->omega, ucell->tpiba, chr);
-        }
-    }
-
     if (use_libxc)
     {
 #ifdef USE_LIBXC