Build native Windows packages for both RDNA4 targets

Add explicit gfx1200/gfx1201 builds, runtime selection and per-target
code objects. Guard staging, packaging and installation against mixed
architectures; retain gfx1201 compatibility and strict K/M arithmetic.

Adapt realdody's target configuration and MSVC binary16/BOOL portability
fixes from ae530b0. Do not import its forced gfx12 layout on RDNA3.
Add exhaustive binary16 and C++/PowerShell target contract regressions.

Validation: 20 CTest gates on RX 9070 XT; four software-only gfx1200 gates;
MSVC portability compile checks; exact 4K K async and M control outputs;
isolated package install/restore and mixed-target rejection tests.
No physical gfx1200 validation is claimed.

Co-authored-by: realdody <[email protected]>

tools/windows/gpu_target.h

@@ -0,0 +1,41 @@
+#pragma once
+#include <array>
+#include <filesystem>
+#include <fstream>
+#include <stdexcept>
+#include <string>
+
+#ifndef D4R_TARGET_ARCH
+#define D4R_TARGET_ARCH "gfx1201"
+#endif
+
+namespace d4r::diag {
+inline bool supported_gpu_target(const std::string& target) {
+    return target == "gfx1200" || target == "gfx1201";
+}
+inline std::string gpu_architecture(const std::string& name) {
+    return name.substr(0, name.find(':'));
+}
+inline void require_gpu_target(const std::string& actual, const std::string& expected) {
+    if (!supported_gpu_target(expected) || gpu_architecture(actual) != expected)
+        throw std::runtime_error("GPU architecture mismatch: detected " + actual + ", package requires " + expected);
+}
+inline std::string code_object_target(const std::filesystem::path& path) {
+    // AMDGPU ELF e_flags machine IDs from LLVM's documented code-object ABI.
+    // Never infer compatibility from a filename or rewrite the ELF target.
+    std::ifstream file(path, std::ios::binary);
+    std::array<unsigned char, 64> header{};
+    if (!file.read(reinterpret_cast<char*>(header.data()), header.size()) ||
+        header[0] != 0x7f || header[1] != 'E' || header[2] != 'L' || header[3] != 'F' ||
+        header[4] != 2 || header[5] != 1 || header[18] != 0xe0 || header[19] != 0)
+        throw std::runtime_error("Invalid AMDGPU ELF64 code object: " + path.string());
+    switch (header[48]) {
+        case 0x48: return "gfx1200";
+        case 0x4e: return "gfx1201";
+        default: throw std::runtime_error("Unsupported AMDGPU code-object target: " + path.string());
+    }
+}
+inline void require_code_object_target(const std::filesystem::path& path, const std::string& expected) {
+    require_gpu_target(code_object_target(path), expected);
+}
+}