perf(trino): 43/43 ClickBench on TestingTrinoServer Path B engine

All 43 ClickBench queries pass on the new TestingTrinoServer-based
engine (replacing DistributedQueryRunner).

Results: 43/43, 431s total, 10.0s avg (single-node, 32g heap, Hive
Parquet with 100M rows from Spark-cleaned data)

Changes:
- Increased query.max-execution-time to 60m (was 10m, blocked CTAS)
- Data: Spark rewrites hits.parquet stripping DATE/TIMESTAMP Parquet
  logical type annotations that Trino 440 Hive reader can't handle.
  EventDate stored as INTEGER (epoch days), EventTime as BIGINT
  (epoch microseconds).

Split-level distribution proven on 3-node TPC-H earlier in session.

Signed-off-by: Peng Huo <penghuo@gmail.com>

Author: penghuo

trino/trino-opensearch/src/main/java/org/opensearch/sql/trino/bootstrap/TrinoEngine.java

@@ -103,7 +103,7 @@ public TrinoEngine(String icebergWarehouse) {
                       "memory.heap-headroom-per-node", headroomStr,
                       "spill-enabled", "false",
                       "exchange.compression-codec", "LZ4",
-                      "query.max-execution-time", "10m"))
+                      "query.max-execution-time", "60m"))
               .build();
 
       // Install connector plugins