feat: add restricted rVV10 plane-wave support

[CleverPhysician]

Linked Issue

• No linked issue; this Draft PR is submitted for maintainer scope and architecture review.

What's changed?

This PR adds an ABACUS-native rVV10 nonlocal-correlation component for the CPU/double/full-complex plane-wave Kohn–Sham SCF path.

• dft_functional selects the semilocal XC term; xc_nonlocal=rvv10 registers an independent additive PotRvv10 component.
• rvv10_b and rvv10_c are validated input parameters and are passed explicitly through setup_estate_pw -> Potential -> PotRvv10.
• The default b=6.3, C=0.0093 reproduces the initial RPW86+PBE parameterization.
• Existing ABACUS PW_Basis FFTs, XC_Functional::grad_rho/grad_dot, charge layouts and pool reductions are reused; QE is used only as a numerical reference.
• MPI pool-distributed PW transforms and collinear nspin=1/2 are supported. For nspin=2, rVV10 uses the total valence-plus-core density and adds the same scalar nonlocal potential to both spin channels.
• Functionals that already contain vdW/VV10 terms are rejected to prevent double counting.

Scope

Supported in this Draft: CPU, double precision, full-complex PW, norm-conserving pseudopotentials, SCF, serial or MPI pool distribution, and nspin=1/2.

Explicitly deferred: GPU, LCAO, USPP/PAW, gamma-only/reduced-gamma layouts, noncollinear nspin=4, forces, stress, relaxation/MD, and simultaneous pairwise D2/D3/D4 corrections. A pre-generated universal kernel table and additional functional-specific parameter sets are follow-up work; the current fixed table is generated once per rVV10 component and validated against the pinned numerical convention.

Unit Tests and/or Case Tests for my changes

Local checks completed after rebasing onto origin/develop 1202e6f:

• python3 -m unittest discover -s tests/rvv10 -p 'test_*.py' — 16/16 passed.
• Python syntax checks passed.
• clang++ -std=c++11 -fsyntax-only passed for the modified input parser and rVV10 core evaluator.
• git diff --check passed.
• agent_governance_check.py --base origin/develop --head HEAD reports no errors; only allowed header-dependency review warnings remain.
• Regression driver includes serial/MPI nspin=1/2 SCF modes, invalid-parameter checks, PBE+rVV10 parser acceptance, vdW/VV10 double-counting rejection, and no-Libxc refusal.

Hosted C++/SCF CI is still required for the full FFT/Libxc/MPI build. This PR remains Draft until the hosted CI and QE/ABACUS energy, potential, spin and MPI comparisons are reviewed by the maintainer.

[Growl1234]

From the first glance I am 👎 at how it's handled:

• dft_functional RVV10 looks like the wrong abstraction. rVV10 is a nonlocal correlation component, so coupling it directly to rPW86+PBE makes SCAN+rVVV10, r2SCAN+rVV10, etc. unnecessarily awkward. The semilocal XC and nonlocal correlation should be represented independently.
• I also don't see much value in merging such a narrowly restricted intermediate implementation. At least MPI and ordinary spin support look like natural parts of the initial feature.
• Not really blocking but worth consideration: Run-time kernel generation also seems unnecessary for a fixed, universal kernel. Pre-generating and storing the kernel data would avoid repeated start-up work (faster) and keep the production path simpler.