fix(linked): support readelf without demangling

[voltjia]

Summary

• Inspect dynamic symbols with readelf --dyn-syms --wide.
• Demangle raw names in one GNU- or LLVM-compatible c++filt process.
• Preserve exact ABI matching across independent nm and readelf symbol inventories.

Motivation

Some supported vendor environments provide a readelf version without the optional --demangle flag. Batch demangling keeps linked-operator discovery portable without weakening its exact-symbol validation.

Depends on #874. N/A - no linked issue.

Type of Change

• feat - new feature / new operator / new platform
• fix - bug fix
• perf - performance improvement (no behavioral change)
• refactor - code restructuring without behavior change
• test - adding or fixing tests only
• docs - documentation only
• build / ci - build system or CI configuration
• chore - tooling, formatting, or other non-code changes
• Breaking change

Platforms Affected

• CPU (WITH_CPU)
• NVIDIA (WITH_NVIDIA)
• Iluvatar (WITH_ILUVATAR)
• MetaX (WITH_METAX)
• Cambricon (WITH_CAMBRICON)
• Moore (WITH_MOORE)
• Ascend (WITH_ASCEND)
• Hygon (WITH_HYGON)
• PyTorch C++ bindings (WITH_TORCH)
• Build system / CMake / CI
• Python bindings / user-facing API

Smoke Test Result

N/A - this PR changes configure-time symbol inspection only; it does not alter an operator runtime path.

Test Results on Supported Platforms

Focused validation at 9de22dd5:

python -m pytest --noconftest tests/test_resolve_linked_ops.py tests/test_generate_wrappers.py -q
39 passed
ruff format --check scripts/resolve_linked_ops.py tests/test_resolve_linked_ops.py
2 files already formatted
ruff check scripts/resolve_linked_ops.py tests/test_resolve_linked_ops.py
All checks passed!

At current head 9de22dd5, build, Ruff, Clang Format, and main/shadow CI for MetaX, Ascend, Cambricon, and Moore pass.

Benchmark / Performance Impact

N/A - configure-time portability fix; runtime paths are unchanged.

Notes for Reviewers

• Exact demangled signatures must appear in both independent symbol inventories.
• c++filt receives raw readelf function symbols in one batch.
• This PR contains no library mapping, operator adapter, or runtime changes.