Allow unparser to override the alias name for the specific dialect

[goldmedal]

Which issue does this PR close?

No corresponding issue at DataFusion. I encountered it in the Wren AI case.

• Fix count syntax for BigQuery Canner/wren-engine#1089

Rationale for this change

Consider the following SQL:

select count(*) from (select 1)

It will be simplified to the following SQL after we move CountWildcardRule to the logical planner #14689

SELECT count(1) AS \"count(*)\" FROM (SELECT 1)

However, some dialects (e.g. BigQuery) don't allow some special characters((, *, @, ...) in an alias name. We should have a way to handle this case for the unparser.

What changes are included in this PR?

• Introduce the interface col_alias_overrides for the unparser dialect.

    /// Allows the dialect to override column alias unparsing if the dialect has specific rules.
    /// Returns None if the default unparsing should be used, or Some(String) if there is
    /// a custom implementation for the alias.
    fn col_alias_overrides(&self, _alias: &str) -> Result<Option<String>> {
        Ok(None)
    }

• Add BigQueryDialect, which will encode the special character in the alias name.

Are these changes tested?

• Add the unit test.
• This change comes from Wren AI fork Bigquery col alias override Canner/datafusion#1 which is used in the production for a while. It works well.

Are there any user-facing changes?

new method for the unparser dialect.

[alamb]

Makes sense to me

[alamb]

it might also help to add a test case here with unparser_dialect GenericDialect too to show the difference in dialects

[alamb]

🚀

[goldmedal]

Thanks @alamb 👍