Extend `nvbench::..._t` aliases to include CUDA specific types

[alliepiper]

E.g. float16_t, etc.

[alliepiper]

This should also include the short name, etc, used for compact table outputs, CLI args, etc.