Is your feature request related to a problem or challenge?

Thanks @jayzhan211 for noticing it in #13581 (comment)

Casting of input should be done automatically according to function signature
For example, the execution logic of corr is implemented in float64 type, so when input is int32 or other compatible type, it will automatically convert to Float64

If the casting logic is implemented correctly during planning, we can safely remove the explicit cast, however, if I try to remove it now, the benchmark query mentioned in #13581 won't run

Describe the solution you'd like

1. Remove cast and fix the error
2. Add the regression test

Describe alternatives you've considered

No response

Additional context

No response