platform.machine() returns the process architecture, so an x64 Python
running under ARM64 emulation reports "AMD64" — the project venv hits
this because torch has no Windows-ARM64 wheel. PROCESSOR_ARCHITEW6432
is empty on Snapdragon X (Prism emulation does not set it), so detect
the native machine via IsWow64Process2 and fall back to platform.machine().

- Document that _IMAGE_FILE_MACHINE_TO_NAME is intentionally limited to
  the host architectures Windows 11 ships on (no 32-bit ARM, no IA64);
  unmapped values fall through to None and log at debug level.
- Log ctypes.get_last_error() when IsWow64Process2 returns 0 so a
  regression does not silently fall back to platform.machine().
- Hoist kernel32 + IsWow64Process2 to module scope, gated by
  sys.platform == "win32", so LoadLibrary runs once per process and
  non-Windows callers don't hit a ctypes.WinDLL error path.
- Add an explicit `if sys.platform != "win32": return None` guard in
  _get_windows_native_machine to match the pattern used in
  _preload_bundled_onnxruntime_dll (#689).
- Split the raw Win32 query (_query_native_machine_via_win32) from the
  name mapping (_get_windows_native_machine) and parametrize tests over
  IMAGE_FILE_MACHINE_* integers so a rename of "ARM64" → "Arm64" in the
  table fails the test instead of silently shipping.

CodeQL flagged "Module is imported with 'import' and 'import from'"
because the prior commit mixed `import ctypes as _ctypes` with
`from ctypes import wintypes as _wintypes`. Hoist `import ctypes` and
`import ctypes.wintypes` to the top of the module and reference the
qualified names; drop the function-local re-import of ctypes.