Support scalarvalue APIs for f16 regarding Pi values

[Jefffrey]

See these four functions:

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

Lines 1342 to 1353 in 33ac70d

	/// Returns a [`ScalarValue`] representing PI's upper bound
	pub fn new_pi_upper(datatype: &DataType) -> Result<ScalarValue> {
	match datatype {
	// TODO: half::f16 doesn't seem to have equivalent
	// https://github.com/apache/datafusion/issues/19465
	DataType::Float32 => Ok(ScalarValue::from(consts::PI_UPPER_F32)),
	DataType::Float64 => Ok(ScalarValue::from(consts::PI_UPPER_F64)),
	_ => {
	_internal_err!("PI_UPPER is not supported for data type: {}", datatype)
	}
	}
	}

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

Lines 1355 to 1366 in 33ac70d

	/// Returns a [`ScalarValue`] representing -PI's lower bound
	pub fn new_negative_pi_lower(datatype: &DataType) -> Result<ScalarValue> {
	match datatype {
	// TODO: half::f16 doesn't seem to have equivalent
	// https://github.com/apache/datafusion/issues/19465
	DataType::Float32 => Ok(ScalarValue::from(consts::NEGATIVE_PI_LOWER_F32)),
	DataType::Float64 => Ok(ScalarValue::from(consts::NEGATIVE_PI_LOWER_F64)),
	_ => {
	_internal_err!("-PI_LOWER is not supported for data type: {}", datatype)
	}
	}
	}

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

Lines 1368 to 1379 in 33ac70d

	/// Returns a [`ScalarValue`] representing FRAC_PI_2's upper bound
	pub fn new_frac_pi_2_upper(datatype: &DataType) -> Result<ScalarValue> {
	match datatype {
	// TODO: half::f16 doesn't seem to have equivalent
	// https://github.com/apache/datafusion/issues/19465
	DataType::Float32 => Ok(ScalarValue::from(consts::FRAC_PI_2_UPPER_F32)),
	DataType::Float64 => Ok(ScalarValue::from(consts::FRAC_PI_2_UPPER_F64)),
	_ => {
	_internal_err!("PI_UPPER/2 is not supported for data type: {}", datatype)
	}
	}
	}

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

Lines 1381 to 1396 in 33ac70d

	// Returns a [`ScalarValue`] representing FRAC_PI_2's lower bound
	pub fn new_neg_frac_pi_2_lower(datatype: &DataType) -> Result<ScalarValue> {
	match datatype {
	// TODO: half::f16 doesn't seem to have equivalent
	// https://github.com/apache/datafusion/issues/19465
	DataType::Float32 => {
	Ok(ScalarValue::from(consts::NEGATIVE_FRAC_PI_2_LOWER_F32))
	}
	DataType::Float64 => {
	Ok(ScalarValue::from(consts::NEGATIVE_FRAC_PI_2_LOWER_F64))
	}
	_ => {
	_internal_err!("-PI/2_LOWER is not supported for data type: {}", datatype)
	}
	}
	}

We should look for a solution that does not involve casting the f32 constant to a f16.

[codetyri0n]

take