zehn_loader: Tell the compiler that the malloced header is aligned

It doesn't know that the function call returns an aligned pointer, so tell
it explicitly. This way it uses aligned loads instead of multiple byte loads.

Here the size goes from 616 to 484, saving more than 20%!

ndless-sdk/tools/zehn_loader/loader.cpp

@@ -72,9 +72,9 @@ int main(int argc, char **argv)
 
 		old_storage = reinterpret_cast<const uint8_t*>(header);
 		syscall<e_memcpy, void*>(mallocd, old_storage, header->file_size);
-		header = reinterpret_cast<const Zehn_header*>(mallocd);
+		header = reinterpret_cast<const Zehn_header*>(__builtin_assume_aligned(mallocd, 4));
 	}
-	
+
 	const Zehn_reloc *reloc_table = reinterpret_cast<const Zehn_reloc*>(reinterpret_cast<const uint8_t*>(header) + sizeof(Zehn_header));
 	//This loader doesn't parse the flag table
 	const uint8_t *flag_table = reinterpret_cast<const uint8_t*>(reloc_table) + sizeof(Zehn_reloc) * header->reloc_count;