Add more support for ScalarValue::Float16 where Float32 and Float64 are supported

[LorrensP-2158466]

Is your feature request related to a problem or challenge?

When fixing #11076, I noticed that not every method in ScalarValue supports a ScalarValue::Float16 while supporting Float32 and Float64:
For example in (from #11076):

datafusion/datafusion/common/src/scalar/mod.rs

Line 1170 in 3051d19

pub fn arithmetic_negate(&self) -> Result<Self> {

We see that F32 and F64 are supported, but not F16

datafusion/datafusion/common/src/scalar/mod.rs

Lines 1178 to 1179 in 3051d19

	ScalarValue::Float64(Some(v)) => Ok(ScalarValue::Float64(Some(-v))),
	ScalarValue::Float32(Some(v)) => Ok(ScalarValue::Float32(Some(-v))),

Describe the solution you'd like

• Every method that supports Float32 and Float64 should support Float16 (including constructors)
• implement From<f16> for ScalarValue
• Add tests
• More to do?

half::f16 supports many of these operations, so I think it is doable. ArrowNativeType and ArrowNativeTypeOp are also implemented so we can use a lot of functions from that trait.

Describe alternatives you've considered

Maybe show in docs why Float16 is not supported in some cases?

Additional context

Can this be a "good first issue" ?
No response

[Lordworms]

I'll take it since it is an issue related to me in #10763