Interesting approach. If I understand your library correctly, you need to choose a bitsize at compile-time? I thought you often wanted unlimited precision (aka. possibly heap-allocated) when using bigints?
Indeed, in my case I didn't need arbitrary precision, just extending uint64 to uint128, uint256, uint512, ... Fixed size are perfectly fine for a variety of use cases, can be stack-allocated and are much faster due to the specialization.
