fix: allocate 4x uom_array/uom_save size for nspin=4

Root cause: In dftu.cpp::init(), pot_index was not multiplied for nspin=4.
This caused eff_pot_pw, uom_array, and uom_save to be allocated with size
for nspin=1, while cal_occ_pw and copy_locale tried to write 4x that amount.

Fix: Add 'else if (nspin == 4) pot_index *= 4;' to match the interleaved
spinor storage requirement.

Also confirmed MPI-2 runs for 223/224/265 are stable (no crashes).

Author: dyzheng

source/source_lcao/module_dftu/dftu.cpp

@@ -170,7 +170,9 @@ void Plus_U::init(UnitCell& cell, // unitcell class
         }
     }
     // allocate memory for eff_pot_pw
-    if(PARAM.inp.nspin == 2) pot_index *= 2; // for spin polarized case, we need to double the size
+    if (PARAM.inp.nspin == 2) pot_index *= 2; // for spin polarized case, we need to double the size
+    else if (PARAM.inp.nspin == 4) pot_index *= 4; // for noncollinear case, need 4x size for interleaved spinors
+
     this->eff_pot_pw.resize(pot_index, 0.0);
     this->uom_array.resize(pot_index, 0.0);
     this->uom_save.resize(pot_index, 0.0);