port(EterLib): Mutex/Thread/FileLoaderThread/GrpVertexBufferDynamic verbatim over Win32Crt kernel-object shims

Adds CreateSemaphore/ReleaseSemaphore/WaitForSingleObject/CloseHandle/
SetThreadPriority/_beginthreadex to Win32Crt; port.file_loader_thread covers
the shims, CThread and both the un-Created (40250 live) and Created loader.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

src/port/common/Win32Crt.cpp

@@ -3,6 +3,10 @@
 #if !defined(_WIN32)
 #include <cctype>
 #include <chrono>
+#include <atomic>
+#include <condition_variable>
+#include <system_error>
+#include <memory>
 #include <cstdio>
 #include <cstring>
 #include <mutex>
@@ -170,6 +174,138 @@ void Sleep(DWORD milliseconds)
 	std::this_thread::sleep_for(std::chrono::milliseconds(milliseconds));
 }
 
+// A HANDLE from CreateSemaphore or _beginthreadex points at one of these. The waitable state is shared
+// with a running thread, so CloseHandle on a live thread only drops the handle, as on Win32.
+namespace
+{
+struct KernelObject
+{
+	virtual ~KernelObject() = default;
+	virtual DWORD Wait(DWORD milliseconds) = 0;
+};
+
+template <class Ready>
+DWORD wait_until(std::unique_lock<std::mutex>& lock, std::condition_variable& cv, DWORD milliseconds, Ready ready)
+{
+	if (milliseconds == INFINITE)
+		cv.wait(lock, ready);
+	else if (!cv.wait_for(lock, std::chrono::milliseconds(milliseconds), ready))
+		return WAIT_TIMEOUT;
+	return WAIT_OBJECT_0;
+}
+
+struct Semaphore final : KernelObject
+{
+	std::mutex mutex;
+	std::condition_variable cv;
+	LONG count = 0;
+	LONG maximum = 0;
+
+	DWORD Wait(DWORD milliseconds) override
+	{
+		std::unique_lock<std::mutex> lock(mutex);
+		const DWORD result = wait_until(lock, cv, milliseconds, [this] { return count > 0; });
+		if (result == WAIT_OBJECT_0)
+			--count;
+		return result;
+	}
+};
+
+struct ThreadState
+{
+	std::mutex mutex;
+	std::condition_variable cv;
+	bool done = false;
+};
+
+struct Thread final : KernelObject
+{
+	std::shared_ptr<ThreadState> state = std::make_shared<ThreadState>();
+
+	DWORD Wait(DWORD milliseconds) override
+	{
+		std::unique_lock<std::mutex> lock(state->mutex);
+		return wait_until(lock, state->cv, milliseconds, [this] { return state->done; });
+	}
+};
+}
+
+HANDLE CreateSemaphore(void*, LONG initial_count, LONG maximum_count, LPCSTR)
+{
+	if (maximum_count <= 0 || initial_count < 0 || initial_count > maximum_count)
+		return NULL;
+	Semaphore* semaphore = new Semaphore;
+	semaphore->count = initial_count;
+	semaphore->maximum = maximum_count;
+	return static_cast<KernelObject*>(semaphore);
+}
+
+// Fails without changing the count when the release would pass the maximum (ERROR_TOO_MANY_POSTS).
+BOOL ReleaseSemaphore(HANDLE handle, LONG release_count, LONG* previous_count)
+{
+	Semaphore* semaphore = handle ? dynamic_cast<Semaphore*>(static_cast<KernelObject*>(handle)) : nullptr;
+	if (!semaphore || release_count <= 0)
+		return FALSE;
+	{
+		std::lock_guard<std::mutex> lock(semaphore->mutex);
+		if (release_count > semaphore->maximum - semaphore->count)
+			return FALSE;
+		if (previous_count)
+			*previous_count = semaphore->count;
+		semaphore->count += release_count;
+	}
+	semaphore->cv.notify_all();
+	return TRUE;
+}
+
+DWORD WaitForSingleObject(HANDLE handle, DWORD milliseconds)
+{
+	if (!handle || handle == INVALID_HANDLE_VALUE)
+		return WAIT_FAILED;
+	return static_cast<KernelObject*>(handle)->Wait(milliseconds);
+}
+
+BOOL CloseHandle(HANDLE handle)
+{
+	if (!handle || handle == INVALID_HANDLE_VALUE)
+		return FALSE;
+	delete static_cast<KernelObject*>(handle);
+	return TRUE;
+}
+
+BOOL SetThreadPriority(HANDLE handle, int)
+{
+	return handle ? TRUE : FALSE;
+}
+
+uintptr_t _beginthreadex(void*, unsigned, unsigned (*start_address)(void*), void* arglist, unsigned, unsigned* thrdaddr)
+{
+	Thread* thread = new Thread;
+	std::shared_ptr<ThreadState> state = thread->state;
+	try
+	{
+		std::thread([state, start_address, arglist] {
+			start_address(arglist);
+			{
+				std::lock_guard<std::mutex> lock(state->mutex);
+				state->done = true;
+			}
+			state->cv.notify_all();
+		}).detach();
+	}
+	catch (const std::system_error&)
+	{
+		delete thread;
+		return 0;
+	}
+	if (thrdaddr)
+	{
+		static std::atomic<unsigned> next_id{1};
+		*thrdaddr = next_id++;
+	}
+	return reinterpret_cast<uintptr_t>(static_cast<KernelObject*>(thread));
+}
+
 char* _strlwr(char* s)
 {
 	for (char* p = s; *p; ++p)