[C++] Make `float16` part of `FloatingPointTypes()`

### Describe the enhancement requested

`FloatingPointTypes()` currently contains only `float32()` and `float64()`, so `float16` is also absent from `NumericTypes()` and `PrimitiveTypes()`.

As a result, every kernel registration loop and test suite that iterates over those lists skips `float16`, and support for it has to be added type by type instead. Several kernel families explicitly reject `HalfFloatType` with `Status::NotImplemented` today.

`float16` should probably be a member of `FloatingPointTypes()`, consistent with the other floating point types.

### Component(s)

C++

Start by locating the C++ definitions of FloatingPointTypes(), NumericTypes(), and PrimitiveTypes(), then inspect the kernel registration loops and test suites that iterate over them. Check how HalfFloatType is handled where kernels return Status::NotImplemented. Done means float16 is included consistently and the affected registrations and tests cover it without requiring type-by-type additions.

Written by the indexing model from the issue text.