RFC: move TypeConversion Checker into cast_to()

[jackwener]

Describe the bug

Current coerce_type mixed with too many functions.
I want to split check type conversion from coerce_type.

such as code

type_coercion.rs:566

// Support the `IsTrue` `IsNotTrue` `IsFalse` `IsNotFalse` type coercion.
// The above op will be rewrite to the binary op when creating the physical op.
fn get_casted_expr_for_bool_op(expr: &Expr, schema: &DFSchemaRef) -> Result<Expr> {
    let left_type = expr.get_type(schema)?;
    let right_type = DataType::Boolean;
    let coerced_type = coerce_types(&left_type, &Operator::IsDistinctFrom, &right_type)?;
    expr.clone().cast_to(&coerced_type, schema)
}

In above code, coerce_types() is just for checking TypeConversion, because we already know coerced_type is Boolean.

---

I think we can check TypeConversion in expr.cast_to().

Here is SQLServer Data type conversion Doc.
https://learn.microsoft.com/en-us/sql/t-sql/data-types/data-type-conversion-database-engine?view=sql-server-2017

To Reproduce

No response

Expected behavior

No response

Additional context

No response

[alamb]

Splitting the "what is the result type of this expression" calculation from "what types should the inputs be coerced to to execute it" is a great idea in my opinion.

I also think it would make the code much easier to reason about if the input arguments were coerced once (during type coercion in the analysis phase) and then never again. Right now I can never quite tell if the inputs to an expr have been coerced or not

[jackwener]

Right now I can never quite tell if the inputs to an expr have been coerced or not

Agree with it, because we should get_type() just for expr have been coerced

[alamb]

I think this issue may be fixed