[SPARK-52899][SQL] Fix QueryExecutionErrorsSuite test to register H2Dialect back

### What changes were proposed in this pull request?
Fixed `QueryExecutionErrorsSuite`'s test `UNRECOGNIZED_SQL_TYPE: unrecognized SQL type DATALINK`. The test is unregistering the H2 dialect and it never registers it back. Suites that rely on H2 can fail because the dialect would be missing.

### Why are the changes needed?
Example suite that fails because of this: https://github.com/PetarVasiljevic-DB/spark/actions/runs/16399646273/job/46337197677

### Does this PR introduce _any_ user-facing change?
No. It is test only change.

### How was this patch tested?

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

Closes apache#51581 from PetarVasiljevic-DB/register_h2_dialect_back.

Authored-by: Petar Vasiljevic <[email protected]>
Signed-off-by: Max Gekk <[email protected]>

Author: PetarVasiljevic-DB

sql/core/src/test/scala/org/apache/spark/sql/errors/QueryExecutionErrorsSuite.scala

@@ -656,6 +656,7 @@ class QueryExecutionErrorsSuite
       sqlState = "42704")
 
     JdbcDialects.unregisterDialect(testH2DialectUnrecognizedSQLType)
+    JdbcDialects.registerDialect(existH2Dialect)
   }
 
   test("INVALID_BUCKET_FILE: error if there exists any malformed bucket files") {