Databricks allows omitting the AS keyword in CTE definitions:
`WITH cte (SELECT ...) SELECT * FROM cte`

Add `supports_cte_without_as()` dialect method and enable it for
Databricks and Generic dialects.

GROUPING SETS used parse_tuple(false, true) which required each
element to be parenthesized, while CUBE and ROLLUP already used
parse_tuple(true, true) allowing bare columns. This inconsistency
meant GROUPING SETS (a, b, c) failed to parse despite being valid
syntax in PostgreSQL, Databricks, and other dialects.

Spark SQL grammar allows parenthesized identifier lists as SELECT
item aliases:

  namedExpression: expression (AS? (identifier | identifierList))?
  identifierList: '(' identifier (',' identifier)* ')'

This enables syntax like:
  SELECT stack(2, 'a', 'b', 'c', 'd') AS (col1, col2)

Databricks, built on Spark SQL, allows identifiers that start with
digits. The Spark SQL ANTLR lexer defines IDENTIFIER as
(UNICODE_LETTER | DIGIT | '_')+, with no restriction on the first
character.

Enable the existing supports_numeric_prefix() for DatabricksDialect.

…fier'