[FEA] Add ssize() members to device_buffer/device_uvector that returns a signed integer

[jrhemstad]

Mixing signed and unsigned types cause annoying compiler warnings necessitating a lot of casts. A common source of this problem is from the fact that size() functions return size_t which is unsigned.

C++20 addressed this by adding a signed size function ssize.

See https://stackoverflow.com/questions/56217283/why-is-stdssize-introduced-in-c20

We should add

int64_t ssize() const noexcept{
   return static_cast<int64_t>(size());
}

to device_buffer and `device_uvector

[jrhemstad]

We could maybe add an assert in there for debugging potential overflows.

assert(size() < std::numeric_limits<int64_t>::max());

[jakirkham]

This is also relevant for Python where Py_ssize_t is used a lot for sizes, but is also a signed integer