Create generic floating-point wrapping type for <NumExponentBits, NumMantissaBits>

All the different extended floating point types like __half, bfloat, fp8, fp4, fp6, etc. are basically just a collection of N bits where M bits are dedicated to mantissa, E bits to exponent.

We can lean into this idea by creating an overarching type that is just:
```
floating_point
```
this would dispatch to corresponding built-in or accelerated types for the appropriate number of bits.

### Tasks
- [ ] https://github.com/NVIDIA/cccl/issues/1666
- [ ] https://github.com/NVIDIA/cccl/issues/2181
- [ ] https://github.com/NVIDIA/cccl/issues/2182