[magnusp]

fix(jdbc): return java.sql-shaped DatabaseMetaData results (#156)

Fixes #156

Hrana 3 is the source of truth for transport, values, metadata, batching,
transactions, and error reporting. java.sql is the source of truth for the
JDBC surface. xerial/sqlite-jdbc 3.53.4.0 is the behavioral reference and
conformance foundation, never a copy target.

What changed

• getProcedures, getProcedureColumns, getColumnPrivileges,
  getTablePrivileges, getBestRowIdentifier, getVersionColumns, and
  getUDTs return an empty result set with the java.sql column list through
  the existing synthesized-metadata path, instead of throwing
  SQLFeatureNotSupportedException.
• getColumns synthesizes the sqlite_schema system table and orders rows the
  way getTables does: temporary, system, table, then view, by name.
• getCrossReference returns no row when a non-empty catalog or schema is
  requested, because the server reports neither.
• getPrimaryKeys sorts rows by COLUMN_NAME, matching xerial's order by cn,
  while KEY_SEQ keeps the declared key position.
• NAMED_PRIMARY_KEY also matches a constraint name glued to the keywords,
  as in CONSTRAINT`NamedPk`PRIMARY KEY(...).

Red

New unit cases at base 3de2ad5 with only the test diff applied:

./mvnw -pl libsql-jdbc -Dtest=LibsqlDatabaseMetaDataTest test
Tests run: 53, Failures: 3, Errors: 14

Vendored DBMetaDataTest at the same base, with a temporary copy selected:

./mvnw -pl libsql-conformance -am -Dit.test=DBMetaDataTest verify
Tests run: 56, Failures: 3, Errors: 8, Skipped: 1
getColumns, getCrossReferenceCatalogSchemaWithQuote, columnOrderOfgetPrimaryKeys,
columnOrderOfgetProcedures, columnOrderOfgetProcedurColumns,
columnOrderOfgetColumnPrivileges, columnOrderOfgetTablePrivileges,
columnOrderOfgetBestRowIdentifier, columnOrderOfgetVersionColumns,
columnOrderOfgetUDTs, columnOrderOfgetTables

Green

At head, with the same temporary copy:

./mvnw -pl libsql-conformance -am -Dit.test=DBMetaDataTest verify
Tests run: 56, Failures: 0, Errors: 1, Skipped: 1

The single remaining error is columnOrderOfgetTables at line 1113:
sqld rejects CREATE TEMP TABLE with SQL_PARSE_ERROR. The xerial control run
passes all 56 on the same copy, so this is remote-model divergence, not a driver
gap. The driver does not fake TEMP support; the later selection PR records the
divergence or excludes the case. version is skipped by its compile-option
assumption.

Full lanes:

./mvnw -pl libsql-jdbc test                           Tests run: 336, Failures: 0, Errors: 0
./mvnw -pl libsql-conformance -am verify              Tests run: 123, Failures: 0, Errors: 0, Skipped: 1
./mvnw -pl libsql-conformance -am -Pxerial-control verify  Tests run: 100, Failures: 0, Errors: 0, Skipped: 1
./mvnw test                                           BUILD SUCCESS
./mvnw spotless:check                                 BUILD SUCCESS
./mvnw -pl libsql-jdbc -am package -DskipTests        BUILD SUCCESS
tools/export-upstream --verify                        verified

Perf, three interleaved runs of ./mvnw -pl libsql-jdbc test: base median
2.47 s, head median 2.50 s (+1.2%).

Logs are in /tmp/xerial-parity-i156/. The vendored tree is unchanged.

Files

libsql-jdbc/src/main/java/com/github/magnusp/libsql/jdbc/LibsqlDatabaseMetaData.java,
libsql-jdbc/src/test/java/com/github/magnusp/libsql/jdbc/LibsqlDatabaseMetaDataTest.java.