Describe the bug

When computing partition_statistics during evalutaion - flamegraph shows a lot of time spend in bounds_check() which happens as part of a Column::data_type() call.

Almost all of the time in bounds_check() is also spend in fmt() which suggests that this goes into the error branch:

impl Column {
    fn bounds_check(&self, input_schema: &Schema) -> Result<()> {
        if self.index < input_schema.fields.len() {
            Ok(())
        } else {
            internal_err!(
                "PhysicalExpr Column references column '{}' at index {} (zero-based) but input schema only has {} columns: {:?}",
                self.name,
                self.index,
                input_schema.fields.len(),
                input_schema.fields().iter().map(|f| f.name()).collect::<Vec<_>>()
            )
        }
    }
}

All occurrences that I hand checked from my example were originating from ProjectionExec::partition_statistics()

To Reproduce

Run with RUST_BACKTRACE enabled.

Expected behavior

data_type() method should not trigger bounds_check() to go to an error path for the column.

Additional context

No response