test(base): expand source_base unit-test coverage - #7888

Reminder

• I have read AGENTS.md and docs/developers_guide/agent_governance.md.
• I have linked an issue or explained why this PR does not need one.
• I have added adequate unit tests and/or case tests, or explained why not.
• I have listed the exact verification commands run and their results.
• I have described user-visible behavior changes, including INPUT parameter changes.
• I have explained core-module impact for ESolver, HSolver, ElecState, Hamilt, Operator, Psi, or other source/ changes.
• I have requested any needed governance exception below.

Linked Issue

No linked issue. This is a focused source_base coverage and decoupling follow-up.

Unit Tests and/or Case Tests for my changes

• Commands run for the original coverage work:
  • OMP_NUM_THREADS=1 ctest --test-dir /tmp/abacus-source-base-coverage --output-on-failure --timeout 1700 -R "^MODULE_BASE"
  • OMP_NUM_THREADS=1 ctest --test-dir /tmp/abacus-source-base-coverage --output-on-failure -R "^MODULE_IO_(output_test|orb_io_test|orb_io_test_parallel|write_wfc_nao|read_exit_file_test)$"
  • OMP_NUM_THREADS=1 ctest --test-dir /tmp/abacus-source-base-coverage --output-on-failure -R "^MODULE_BASE_(parallel_reduce_test|parallel_device|parallel_device_parallel)$"
  • lcov capture/extract/remove over /tmp/abacus-source-base-coverage, restricted to production files under source/source_base.
• Commands run after syncing to upstream develop at c2fa3ac19:
  • cmake -S . -B build-pr7888 -G Ninja -DBUILD_TESTING=ON -DENABLE_MPI=ON -DCMAKE_BUILD_TYPE=Debug
  • cmake --build build-pr7888 --target MODULE_BASE_para_gemm MODULE_BASE_parallel_device MODULE_BASE_parallel_domain_grid -j4
  • OMP_NUM_THREADS=1 ctest --test-dir build-pr7888 --output-on-failure -R '^MODULE_BASE_(para_gemm|para_gemm_parallel|parallel_device|parallel_device_parallel|parallel_domain_grid|parallel_domain_grid_parallel)$'
  • OMP_NUM_THREADS=1 ctest --test-dir build-pr7888 --output-on-failure -R '^MODULE_BASE' -j4
  • clang-format --dry-run --Werror source/source_base/test_parallel/parallel_device_test.cpp source/source_base/test_parallel/parallel_domain_grid_test.cpp source/source_base/test_parallel/test_para_gemm.cpp
  • python3 tools/03_code_analysis/agent_governance_check.py --base upstream/develop --head HEAD --format text
  • python3 tools/03_code_analysis/code_quality_score.py $(git diff --name-only upstream/develop...HEAD | grep -E '\.(cpp|h)$')
  • git diff --check upstream/develop...HEAD
• Result summary:
  • MODULE_BASE: 77/77 passed against upstream develop at c2fa3ac19, including all affected serial and four-process MPI tests.
  • Related MODULE_IO tests: 5/5 passed in the original coverage verification, including four-process orbital IO.
  • MODULE_BASE_LIBM_UTs passed in a separate ENABLE_ABACUS_LIBM=ON targeted configuration.
  • Final CPU/MPI production coverage from the original coverage run: 87.7% lines (15430/17598), 97.5% functions (1835/1883).
  • No compiled production source file remains entirely at zero function coverage.
  • Code-quality scoring: 27/27 changed C++ files passed the 60-point threshold (average 89.0).
  • Governance check has no deterministic errors; its documentation warning is addressed below.
• Checks not run, with reason:
  • CUDA, ROCm, and DSP runtime paths were not executed locally because those accelerator environments are unavailable in this CPU/OpenMPI build. CUDA compilation is covered by CI.
  • The remaining 48 gcov zero-count symbols are 40 macro-generated linalg dispatch lambdas for invalid scalar/tensor type combinations plus 8 compiler-generated destructor/inline duplicates or abort-only handlers. The public behavior behind these symbols is covered; aborting child processes cannot flush gcov counters.
  • A whole-file clang-format --dry-run --Werror over every touched legacy file reports existing formatting differences. New standalone tests and directly formatted files pass targeted formatting checks; this PR intentionally avoids unrelated repository-wide formatting churn.
  • An all-target local build was not completed because it would compile roughly 3000 unrelated targets. All MODULE_BASE targets were built, installed with their test assets, and the complete MODULE_BASE test selection passed.

What changed?

• Added focused coverage for numerical utilities, BLAS/LAPACK wrappers, container integer kernels, MPI communication and reduction type variants, parallel grids/domains, para_gemm float variants, CSR complex parsing, and remaining CPU helpers.
• Added reusable MPI test setup and new focused test targets where needed.
• Fixed the no-accelerator set_memory<T, DEVICE_GPU> stub signature to match its declared interface.
• Fixed TensorBuffer move ownership transfer to clear the moved-from allocator and prevent double deletion; also keeps allocated-byte accounting correct after resize/copy/move assignment.
• Replaced two backend-sensitive exact floating-point comparisons with tight tolerances.
• Synced the PR branch with current upstream develop and migrated the added parallel tests from the removed communication_domain.h/size() API to parallel_cell.h and the existing MPICommGroup wrapper.

Governance Notes

• INPUT/docs changes: none. This PR changes tests and internal source_base correctness only; no user-facing INPUT behavior or parameter metadata changes, so docs/parameters.yaml and docs/advanced/input_files/input-main.md do not require updates.
• Core module impact: limited to source_base tests plus two internal correctness fixes. No new GlobalV, GlobalC, or PARAM dependency is introduced, and source_base remains independently testable.
• The parallel test compatibility fix uses the existing MPICommGroup wrapper and adds no direct MPI routine calls.
• Exceptions requested: none.