NVCC's `std::construct_at` requires those default constructors

* So far the results may be: * * - `i32_addition` - @b 0ns, as the compiler optimized the code away. * - `i32_addition_inline_asm` - @b 0.2ns, a

* It's common knowledge that floating-point math can be costly, but even * integer operations can be surprisingly expensive. @b Division and modulo

* Division takes around ~10 CPU cycles or @b 2.5ns. However, if the divisor * is known at compile time, the compiler can replace the division with

* An important optimization trick is that 32-bit integer division can be * performed using 64-bit double-precision floating-point division. This *

clang-format off