[WIP] Add Kokkos::Impl::assume_aligned for broader compiler support

[cwpearson]

This is a work in progress

• gcc and clang have __builtin_assume_aligned
• nvcc added assume_aligned in 11.2
• msvc has __assume() which can be used in a similar way
• intel has __assume_aligned

• HIP?
• What are the right macros to use to detect the compiler and host code vs device code?

If MSVC can't see __assume through a function return a pure macro-based approach may be more appropriate.

[cwpearson]

Dropping this, not much progress was made and if we have a good use for it we can redo it.