DeltaP PW: fix multi-k segfault — use Gamma mesh + pw_rho

Two fixes for multi-k Berry phase during SCF:
  1. Use Gamma mesh (not Monkhorst-Pack) — Monkhorst-Pack
     causes segfault for isolated molecules in large boxes.
     Gamma 1x1x2 works correctly.
  2. Pass pw_rho (smooth density grid) instead of pw_rhod
     (dense USPP grid) as rhopw to compute_total_gamma_pw().
     unkdotp_G0 uses rhopw->recip2real which requires
     the smooth FFT grid.

Multi-k verified (H2O, Gamma 1x1x2):
  gamma_total = -0.2302 rad (same as Gamma-only -0.2298)
  per-atom: O=-0.1512, H=-0.0398 (stable across iterations)
  E_tot = -417.089 eV (vs -417.169 eV Gamma-only)

BN LCAO regression passes.

Author: dyzheng

source/source_esolver/esolver_ks_pw.cpp

@@ -283,7 +283,7 @@ void ESolver_KS_PW<T, Device>::iter_finish(UnitCell& ucell, const int istep, int
 
     // DeltaP: compute gamma and update lambda after SCF iteration
     pw_deltap::deltap_iter_finish(ucell, this->drho,
-        this->stp.psi_cpu, this->kv, this->pw_wfc, this->pw_rhod, PARAM.inp);
+        this->stp.psi_cpu, this->kv, this->pw_wfc, this->pw_rho, PARAM.inp);
 
     // the output quantities
     ModuleIO::ctrl_iter_pw(istep, iter, conv_esolver, this->stp.psi_cpu, 

tests/deltap_pw_h2o/INPUT

@@ -27,4 +27,4 @@ deltap_gdir   3
 deltap_lambda_step     0.01
 deltap_lambda_mixing   0.1
 deltap_inner_thr       1e-3
-onsite_radius 6.0
+onsite_radius 6.0