Bump default value of max split count per task for FTE

With limit of 256 splits per task it was possible to end up with very
small tasks and possibly even reach limit of number of tasks per stage
(2^^15). The problem appeared when splits were very small.
Task sizing should be mostly done based on split size; max
number of splits should be treated as extra limit to be reached only in
rare situations. This PR bumps default value 8x to 2048.

core/trino-main/src/main/java/io/trino/execution/QueryManagerConfig.java

@@ -132,7 +132,7 @@ public class QueryManagerConfig
     private int faultTolerantExecutionHashDistributionWriteTaskTargetMaxCount = 2000;
 
     private DataSize faultTolerantExecutionStandardSplitSize = DataSize.of(64, MEGABYTE);
-    private int faultTolerantExecutionMaxTaskSplitCount = 256;
+    private int faultTolerantExecutionMaxTaskSplitCount = 2048;
     private DataSize faultTolerantExecutionTaskDescriptorStorageMaxMemory = DataSize.ofBytes(round(AVAILABLE_HEAP_MEMORY * 0.15));
     private int faultTolerantExecutionMaxPartitionCount = 50;
     private int faultTolerantExecutionMinPartitionCount = 4;

core/trino-main/src/test/java/io/trino/execution/TestQueryManagerConfig.java

@@ -98,7 +98,7 @@ public void testDefaults()
                 .setFaultTolerantExecutionHashDistributionWriteTasksToNodesMinRatio(2.0)
                 .setFaultTolerantExecutionHashDistributionWriteTaskTargetMaxCount(2000)
                 .setFaultTolerantExecutionStandardSplitSize(DataSize.of(64, MEGABYTE))
-                .setFaultTolerantExecutionMaxTaskSplitCount(256)
+                .setFaultTolerantExecutionMaxTaskSplitCount(2048)
                 .setFaultTolerantExecutionTaskDescriptorStorageMaxMemory(DataSize.ofBytes(round(AVAILABLE_HEAP_MEMORY * 0.15)))
                 .setFaultTolerantExecutionMaxPartitionCount(50)
                 .setFaultTolerantExecutionMinPartitionCount(4)

docs/src/main/sphinx/admin/fault-tolerant-execution.md

@@ -237,7 +237,7 @@ properties only apply to a `TASK` retry policy.
     May be overridden for the current session with the
     `fault_tolerant_execution_max_task_split_count` [session
     property](session-properties-definition).
-  - `256`
+  - `2048`
 * - `fault-tolerant-execution-arbitrary-distribution-compute-task-target-size-growth-period`
   - The number of tasks created for any given non-writer stage of arbitrary
     distribution before task size is increased.