[SPARK-58193][SQL] Classify Databricks syntax errors by message

[alekjarmov]

What changes were proposed in this pull request?

Recognize remote Databricks parse errors when the JDBC driver reports the SYNTAX_ERROR condition only in its message.
Optimise isSyntaxErrorBestEffort for Databricks to catch syntax error even if SQLState is missing.

Why are the changes needed?

To improve the error classification.

Does this PR introduce any user-facing change?

No

How was this patch tested?

Added a unit-test and checked a case on a Databricks workspace

Was this patch authored or co-authored using generative AI tooling?

Generated-by: Claude 4.7

[cloud-fan]

0 blocking, 2 non-blocking, 0 nits.
The change is correct and consistent with the MsSqlServerDialect precedent (message-substring classification for a vendor without a reliable syntax-error SQLState); the SQLState-42 path is retained, so the check only broadens. Two non-blocking test suggestions below.

Suggestions (2)

• JDBCTableCatalogSuite.scala:232: pure dialect unit test placed here instead of JDBCSuite where the other DatabricksDialect unit tests live — see inline
• JDBCTableCatalogSuite.scala:233: only the returns-true path is asserted; a negative case would guard the best-effort substring match — see inline

[cloud-fan]

2 addressed, 0 remaining, 0 new. (0 newly introduced, 0 late catches.)
0 blocking, 0 non-blocking, 0 nits.

Verification

Static trace: JDBCRDD consults this hook during schema resolution and query execution, while DatabricksDialect preserves the existing SQLState path and adds only the explicit vendor marker fallback. Tests were not run as part of this review.

PR description suggestions

• Update the user-facing-change section from No: matching this driver error now produces Spark's JDBC_EXTERNAL_ENGINE_SYNTAX_ERROR classification during schema resolution or query execution.