[andygrove]

Spark SQL types a numeric literal by a case-insensitive suffix: Y, S and L after an integer, and F, D and BD after any number. SparkSqlDialect reads every suffix except an uppercase L as a column alias:

SELECT 1D, 1.5BD, 10Y, 1l
-- today: SELECT 1 AS D, 1.5 AS BD, 10 AS Y, 1 AS l

This adds Dialect::supports_numeric_literal_suffixes, enabled for Spark. The tokenizer then reads Y, S, F, D and BD into a new Token::SuffixedNumber, which parses to Value::SuffixedNumber(String, NumberSuffix). An L in either case still gives Value::Number(_, true). As in Spark, L, S and Y only follow an integer, and a suffix directly followed by an identifier character, as in 1Dx, is not one. Other dialects are unchanged.

Spark docs: https://spark.apache.org/docs/latest/sql-ref-literals.html#numeric-literal

This PR was prepared with AI assistance (Claude).

Dialects: Spark