sync changes with afqmc

several changes to memory, 1: added alloc_XXX functions to memory::allocator, 2: changed to raw_pointer_cast in align function in monotonic allocator, 3: added copy constructor to cuda::allocator and memory::allocator, needed in case stateful allocators are used, 4: added memory resource to cuda namespace based on the existing allocator, 5: made a few changes to cuda::ref to allow r-value binary operations with cuda::ref on both sides, added tests for this also.