source_estate: print_etot takes its INPUT parameters explicitly, removing elecstate_print_test's access hack

[Critsium-xy]

What this does

All fourteen of elecstate_print.cpp's global reads were inside one function,
print_etot(), and it has exactly one production call site. print_etot()
now takes const Input_para& inp plus the derived two_fermi flag, and
elecstate_print.cpp is PARAM-free.

PARAM/GlobalV/GlobalC ledger: added 1, removed 58, net_delta = −57.
Zero blockers.

The two new parameters are inserted before the existing defaulted arguments,
so no new default argument is introduced (governance blocks that).

esolver_ks.cpp already holds an injected inp_ and already passes
*this->inp_ to ModuleIO::write_bands on the same code path, so the single call
site follows a pattern that is already in that file.

Why the whole Input_para rather than separate flags

print_etot() reads thirteen distinct INPUT keys — basis_type,
out_freq_elec, scf_nmax, vdw_method, dft_plus_u, imp_sol,
efield_flag, gate_flag, deepks_scf, ml_exx, out_level, nspin,
esolver_type, ks_solver. Thirteen parameters would be worse than the
aggregate the function is actually reporting: its job is to print the
INPUT-driven state. two_fermi is passed separately because it is derived and
lives in System_para, not Input_para.

Test side

The fixture owns an Input_para and a two_fermi bool instead of writing the
private half of PARAM. Its PARAM.sys.log_file write was dead — no source
compiled by MODULE_ESTATE_elecstate_print reads it — and is dropped.

I also checked the other headers the macro region covered (klist.h,
charge.h, efield.h, gatefield.h, xc_functional.h): the test touches no
private member of any of them. Its two XC_Functional::func_type /
ked_flag lines are out-of-line definitions, which need no access.

Verification

Remote Linux box, cmake -B build -G Ninja -DBUILD_TESTING=ON -DENABLE_LCAO=ON -DENABLE_MPI=ON -DENABLE_OPENMP=ON: configure and build exit 0,
3354/3354 targets, zero errors. agent_governance_check.py: zero blockers,
no access-hack ratchet finding (1 removed, 0 added).

No production logic changed. No assertion or expected value changed. No INPUT
parameter or user-facing behaviour changed, so no documentation update is
required.

The two sibling files are not in this PR

elecstate_energy_test and elecstate_pw_test were scoped alongside this one,
and measuring them showed they are not the same problem, so they are left for
separate PRs rather than bundled in half-done:

• elecstate_energy_test (1 macro) — 7 of its 12 written keys are dead
  (basis_type, esolver_type, lspinorb, nelec, nlocal, nupdown,
  two_fermi have no reader in MODULE_ESTATE_elecstate_energy). The live ones
  need nspin injected into cal_delta_eband (8 call sites) and
  cal_delta_escf (2), and imp_sol / sc_mag_switch / dft_plus_u / nspin
  into cal_energies (9) — 19 call sites across 12 files, and two of them
  (chgmixing.cpp, rdmft.cpp) have no injected inp_ to draw on. Mechanical
  but wide.

• elecstate_pw_test (2 macros) — the hard one. Of its 25 global reads, three
  are in ~ElecStatePW(), and a destructor cannot take parameters. device,
  precision, double_grid and use_uspp therefore have to become member state
  captured at construction — the same move this campaign used for Charge::nelec
  in source_estate: decouple Charge from PARAM, plus two setters on XC_Functional — removes charge_test's two access hacks (touches module_xc) #7990, but across six methods of a template class. That is a design change,
  not a signature change. It also needs the XC_Functional::set_func_type() added
  in source_estate: decouple Charge from PARAM, plus two setters on XC_Functional — removes charge_test's two access hacks (touches module_xc) #7990, so it should land after that.

Continues #7921 / #7949 / #7952 / #7953 / #7963 / #7964 / #7965 / #7966 / #7967 /
#7980 / #7981 / #7982 / #7983 / #7984 / #7987 / #7988 / #7989 / #7990.

🤖 Generated with Claude Code