Relax key/payload/slot type constraints to enable 1B and 2B types

[sleeepyjack]

I think we are too strict on the key/payload/slot type constraints. cuda::atomic_ref supports 1B and 2B operations. This is is to be explored in a future PR.

Originally posted by @sleeepyjack in #803

[ksapru]

This seems like a good first issue for me to explore this repository and learn more. Happy to take this on if it's still up for grabs. cc: @sleeepyjack

[sleeepyjack]

Hi @ksapru,

Yes, this is a good fit for "good first issue." I'm fairly sure our implementation already works with key/value types smaller than 4 bytes (famous last words).

The best way to approach this is to add uint8_t and uint16_t to the type axes in our unit tests. The static_set tests are a good starting point. Once those pass, the rest should fall in line easily, since the other data structures share the same implementation in detail/open_addressing_ref_impl.cuh and detail/open_addressing_impl.cuh.

Let me know if you need any help getting started.

[sleeepyjack]

Next steps after #812 : (1) Add (almost) exhaustive test coverage to all OA-based data structures and (2) extend benchmarks.