RFC: Add `result type` for Expr

### Is your feature request related to a problem or challenge?

In the current case, when we go to get the `result type` of Expr, we don't save it.
It also means that if we get the `result type` multiple times, it may require multiple runtime evaluate.

BTW, this feature is also related with `Type Coercion`, we should/must get `result type` after `Type Coercion`, because it may be invalid (need coerced type).

### Describe the solution you'd like

_No response_

### Describe alternatives you've considered

_No response_

### Additional context

_No response_

The issue provides no file or test names. Start by locating the Expr result-type evaluation and the Type Coercion logic, then trace how often runtime evaluation occurs. Done should include a defined approach for retaining the coerced result type and avoiding repeated evaluation, with tests covering both behaviors.

Written by the indexing model from the issue text.