feat(duckdb): Add transpilation support for FLATTEN (ARRAY_FLATTEN)

[fivetran-amrutabhimsenayachit]

ARRAY_FLATTEN function from Snowflake is not transpiling to DuckDB's FLATTEN function. SQLGlot currently parses ARRAY_FLATTEN as an Anonymous function without mapping it to the internal Flatten expression type, resulting in failed transpilation to DuckDB.

 python3 -c "import sqlglot; print(sqlglot.transpile(\"SELECT ARRAY_FLATTEN([[1, 2, 3], [4], [5, 6]]) AS result\", read='snowflake', write='duckdb')[0])" | duckdb
 DuckDB: SELECT FLATTEN([[1, 2, 3], [4], [5, 6]]) AS result
┌────────────────────┐
│       result       │
│      int32[]       │
├────────────────────┤
│ [1, 2, 3, 4, 5, 6] │
└────────────────────┘

[github-actions]

SQLGlot Integration Test Results

Comparing:

• this branch (sqlglot:RD-1147617-array-flatten, sqlglot version: RD-1147617-array-flatten)
• baseline (main, sqlglot version: 28.6.1.dev82)

⚠️ Limited to dialects: snowflake

By Dialect

dialect	main	sqlglot:RD-1147617-array-flatten	difference	links
snowflake -> duckdb	615/847 passed (72.6%)	615/847 passed (72.6%)	No change	full result / delta
snowflake -> snowflake	847/847 passed (100.0%)	847/847 passed (100.0%)	No change	full result / delta

Overall

main: 1694 total, 1462 passed (pass rate: 86.3%), sqlglot version: 28.6.1.dev82

sqlglot:RD-1147617-array-flatten: 1694 total, 1462 passed (pass rate: 86.3%), sqlglot version: RD-1147617-array-flatten

Difference: No change

[geooo109]

This function exists in various dialects but I think we missing tests, we could add them in a separate PR. cc: @VaggelisD , @georgesittas

[georgesittas]

This function exists in various dialects but I think we missing tests, we could add them in a separate PR. cc: @VaggelisD , @georgesittas

@geooo109 go for it. If we cover all dialects it may be worth renaming Flatten to ArrayFlatten. The former was introduced to support Snowflake's FLATTEN(...) operator, which was later mapped to Explode and hence that context was lost.