Support dataflow tracking that is wider than 64bit

[xusheng6]

Currently we are tracking dataflow using an uint64_t, which works great, but would break if the input is wider than that. Modern compilers, especially MSVC, tend to use XMM registers when it can, and the underlying operation is often times just trivial, i.e., copy, xor, etc. These would all be handled if we support up to 128-, 256-, 512- bit dataflow.

https://github.com/chfast/intx might be a helpful library for the purpose. Also we probably wouldn't need arbitrary precession integer arithmetic -- 512 bit is honestly wide enough for most practical usage scenario

Not super related, but worth linking together: Vector35/debugger#184