Lets assume I want to be able to swap in and out allocators none of
which will be the gc_allocator...
What is best practice for using std.experimental.allocator in @nogc code?