3rdparty(mlas): fix MinGW build of aligned scratch buffer

MlasThreadedBufAlloc() and its ThreadedBufHolder guarded the
_aligned_malloc / _aligned_free path on _MSC_VER. Non-MSVC Windows
toolchains (MinGW, clang) therefore fell through to posix_memalign(),
which the Windows CRT does not provide, so the build failed with
"'posix_memalign' was not declared in this scope".

Guard the aligned-allocation path on _WIN32 instead, in both the holder
declaration (mlasi.h) and definition (platform.cpp) so the unique_ptr
deleter type stays consistent across translation units, and include
<malloc.h> on Windows since MinGW only declares the _aligned_* functions
there. The non-Windows posix_memalign / aligned_alloc fallback is
unchanged.

Fixes #29350

3rdparty/mlas/lib/mlasi.h

@@ -48,6 +48,7 @@ Abstract:
 #endif
 #include <windows.h>
 #include <intrin.h>
+#include <malloc.h>
 #else
 #if defined(__arm__) || defined(__aarch64__)
 #include <arm_neon.h>
@@ -3014,7 +3015,7 @@ MlasReadTimeStampCounter(void)
 
 constexpr size_t ThreadedBufAlignment = 64;
 extern thread_local size_t ThreadedBufSize;
-#ifdef _MSC_VER
+#ifdef _WIN32
 extern thread_local std::unique_ptr<uint8_t, decltype(&_aligned_free)> ThreadedBufHolder;
 #else
 extern thread_local std::unique_ptr<uint8_t, decltype(&free)> ThreadedBufHolder;
@@ -3034,7 +3035,7 @@ void
 MlasThreadedBufAlloc(size_t size)
 {
     if (size > ThreadedBufSize) {
-#ifdef _MSC_VER
+#ifdef _WIN32
         ThreadedBufHolder.reset(
             reinterpret_cast<uint8_t*>(_aligned_malloc(size, ThreadedBufAlignment)));
 #elif (__STDC_VERSION__ >= 201112L) && !defined(__APPLE__)

3rdparty/mlas/lib/platform.cpp

@@ -1088,7 +1088,7 @@ MlasPlatformU8S8Overflow(
 
 #endif
 thread_local size_t ThreadedBufSize = 0;
-#ifdef _MSC_VER
+#ifdef _WIN32
 thread_local std::unique_ptr<uint8_t, decltype(&_aligned_free)> ThreadedBufHolder(nullptr, &_aligned_free);
 #else
 thread_local std::unique_ptr<uint8_t, decltype(&free)> ThreadedBufHolder(nullptr, &free);