Spark, Databricks: parse the << and >> shift operators

[moshap-firebolt]

Spark and Databricks both have the << and >> bitwise shift operators. The parser already supports them behind supports_bitwise_shift_operators; this turns that on for SparkSqlDialect and DatabricksDialect.

Spark's grammar binds a shift tighter than & and looser than +. The shared default puts shifts at the Caret level, which sits below &, so 1 & 3 << 1 would read as (1 & 3) << 1. Both dialects override get_next_precedence to place shifts just above Ampersand.

Tests in tests/sqlparser_spark.rs cover both operators, the operator tag, and the tree shape against & and +. The fork's full suite, fmt and clippy are clean.

---

Note

Low Risk
Dialect-only parser changes with targeted tests; no auth, data, or runtime execution paths affected.

Overview
Spark and Databricks now parse bitwise shift operators << and >> by enabling supports_bitwise_shift_operators on both dialects.

Both dialects override get_next_precedence so shifts bind tighter than & and looser than + (e.g. 1 & 3 << 1 → 1 & (3 << 1)), matching Spark’s grammar instead of the default precedence that would group shifts below &.

Spark tests in test_bitwise_shift_operators cover parsing, operator kinds, and precedence against & and +.

^{Reviewed by Cursor Bugbot for commit b4c500b. Bugbot is set up for automated code reviews on this repo. Configure here.}