Optimize project scanning on Windows

We were using FindFirstFileW instead of FindFirstFileExW, which provides
two optimization opportunities for folder scanning:

- Specifying FindExInfoBasic avoids computing short file names
  (cAlternativeFileName), which we never use
- Specifying FIND_FIRST_EX_LARGE_FETCH increases the internal buffer
  sizes used for reading file entries

When folder hierarchy is cached (hot), the first optimization improves
scan performance by ~10%. When folder hierarchy is read from disk (cold),
the second optimization improves scan performance by ~25%. The numbers
are measured on Windows 11 / Samsung 970 EVO / x64 on Linux kernel
source tree (65k files)

src/fileutil_win.cpp

@@ -46,7 +46,7 @@ static bool traverseDirectoryRec(const wchar_t* path, const char* relpath, const
 	std::wstring query = path + std::wstring(L"/*");
 
 	WIN32_FIND_DATAW data;
-	HANDLE h = FindFirstFileW(query.c_str(), &data);
+	HANDLE h = FindFirstFileExW(query.c_str(), FindExInfoBasic, &data, FindExSearchNameMatch, NULL, FIND_FIRST_EX_LARGE_FETCH);
 
 	if (h == INVALID_HANDLE_VALUE)
 		return false;