Which issue does this PR close?

• Closes #.

Rationale for this change

DataFusion's SQL unparser lacked Snowflake dialect support for UNNEST in the FROM clause. Snowflake does not have a native UNNEST table factor — instead it uses the FLATTEN table function to expand arrays. Without this support, logical plans containing Unnest nodes could not be round-tripped to valid Snowflake SQL.

The Datafusion SQL:

SELECT * FROM UNNEST([1,2,3]) t(a)

It can be transpiled to the equivalent SQL in Snowfalke dialect:

SELECT "t"."a" FROM TABLE(FLATTEN([1, 2, 3], '', false, false, 'ARRAY')) AS "t" ("SEQ", "KEY", "PATH", "INDEX", "a", "THIS")

What changes are included in this PR?

• Added two new Dialect trait methods:
  • unparse_unnest_table_factor — lets a dialect convert an Unnest plan node into a dialect-specific TableFactor (e.g. TABLE(FLATTEN(...)) for Snowflake)
  • relation_alias_overrides — lets a dialect intercept and rewrite the alias applied to a relation builder, used here to patch the VALUE column name in the FLATTEN alias list
• Add SnowflakeDialect with the unparse_unnest_table_factor implementation to show how to use it.
• Fix the quoted identifier issue for the subquery_alias_inner_query_and_columns method.

Are these changes tested?

yes.
Wren AI has been using it in production for a while. Canner#9

Are there any user-facing changes?

yes, new method for the unparser dialect.