source_basis/module_ao: use the getters that already exist instead of #define private public

[Critsium-xy]

Third tranche of the #define private public cleanup, after #7940 (tests
driving global PARAM) and #7949 (tests calling private methods). This one
covers the case where the member the test reaches for already has a public
getter — so the macro comes off with no production change at all.

Three cases, and the middle one is a trap

orb_nonlocal_test.cpp and orb_read_test.cpp reach into Numerical_Nonlocal
and LCAO_Orbitals. Neither file touches PARAM; nothing is added to a
production header. The entire diff is in the two test files.

1. The member is used as a plain value — read it through its getter.
lcao_.kmesh becomes lcao_.get_kmesh() (4 sites); the five SetTypeInfo
assertions become nn.getLabel(), getType(), getLmax(), get_rcut_max(),
get_nproj(). Those already compared against the fixture's inputs, so they stay
real checks.

2. The assertion is EXPECT_EQ(obj.get_x(), obj.x). Substituting the
getter here would produce EXPECT_EQ(get_x(), get_x()) — an assertion that
cannot fail. Mechanically removing the macro this way would leave the test
green and empty, which is worse than the macro. These are re-anchored to the
value the object was actually given.

orb_nonlocal_test already had one line in the right form
(EXPECT_EQ(nn.get_rcut_max(), rcut_max_)); the other four now match it.

In LCAO_Orbitals::Getters, seven of the twelve assertions covered private
members. ntype and lmax are passed straight into Read_Orbitals, so they
anchor to ntype_ / lmax_. The remaining four are derived, and the
production formula is deliberately not restated in the test — a test that
recomputes the thing it is checking passes even when the formula is wrong. They
are asserted as the concrete values this fixture implies, each with its
provenance in a comment:

	value	where it comes from
kmesh	1113	int(sqrt(ecutwfc)/dk) + 4 = int(sqrt(123)/0.01) + 4
nchimax	2	H is 2s1p, O is 2s2p1d
lmax_d, nchimax_d	2	from jle.orb
rcutmax_Phi	8 au	H is 8 au, O is 7 au

Each was measured against the built test, and kmesh additionally cross-checked
by hand against Read_Orbitals so the assertion records the correct value rather
than merely the observed one.

3. The test wrote a private member to build a fixture —
nnl[i].rcut = 1.0 in NumericalNonlocalTest::SetUp.
Numerical_Nonlocal_Lm derives rcut from the last point of its radial mesh
(rcut = r_radial_in[nr-1]), so each projector is now built through the public
set_NL_proj() with a minimal three-point mesh whose endpoint is the wanted
rcut. This removes the write and exercises set_NL_proj, which no assertion in
this file reached before.

ecutwfc, dk, dR, Rmax and dr_uniform are public members of
LCAO_Orbitals and never needed the macro. The four that have a corresponding
fixture input are anchored to it for consistency; dr_uniform is left as it was.

Result

Macro occurrences in these two files: 2 -> 0. No #undef private is added,
and no file whose macro survives is touched.

Verification

Linux, cmake -B build -G Ninja -DBUILD_TESTING=ON -DENABLE_LCAO=ON -DENABLE_MPI=ON -DENABLE_OPENMP=ON,
followed by cmake --install build — which matters here: the module_ao tests
get their orbital data from install(DIRECTORY lcao_H2O ...), so without the
install step ORB_read_test fails and ORB_atomic_lm_test /
ORB_nonlocal_lm_test segfault on missing input, both before and after this
change.

• build: 0 errors (branch and base both BUILD_EXIT=0).
• MODULE_AO_ORB_nonlocal_test, MODULE_AO_ORB_read_test,
  MODULE_AO_ORB_atomic_lm_test, MODULE_AO_ORB_nonlocal_lm_test: all pass.
• full unit suite: 2 of 339 fail — the failure set is identical to the base
  commit (9cbcc0b55) built, installed and run in a parallel worktree in the
  same environment; both comm directions are empty. The two are
  MODULE_HSOLVER_diago_hs_parallel and MODULE_HSOLVER_LCAO, both
  mpirun-based and pre-existing.
• agent_governance_check.py: 0 errors. The access-hack ratchet reports
  nothing (2 removed, 0 added) and no PARAM/GlobalV/GlobalC reference is
  added or removed, so the global-dependency budget is untouched.

No INPUT parameter and no user-visible behaviour changed, so
docs/parameters.yaml and docs/advanced/input_files/input-main.md need no
update.

What is not here

The other two module_ao tests, orb_atomic_lm_test.cpp (887 lines) and
orb_nonlocal_lm_test.cpp (755 lines), hold the bulk of this category — roughly
190 further sites where a member with an existing getter is read directly. They
are left for a separate PR because they cannot be finished the same way: each
also calls private methods (cal_kradial, cal_rradial_sbpool, plot,
freemem, renew, get_kradial) and orb_atomic_lm_test reads psir, which
has no getter. Those need a friend grant in a production header — the #7949
pattern — which is a different review question from "use the accessor that
already exists", and mixing the two would obscure both.

Their getter-vs-member block has the same shape as the one fixed here but over
arrays (get_psi() vs psi), where there is no scalar input to re-anchor to;
that block is better routed through the fixture once the friend is in place.

🤖 Generated with Claude Code

[mohanchen]

LGTM