Load the serialized Dag in ti_run only for stub tasks (#75795)

* Load the serialized Dag in ti_run only for stub tasks

Starting any task deserialized its whole Dag so the route could find out
whether the task was a stub and read its arg_bindings. Denormalize the
serialized is_stub flag onto task_instance, which ti_run already fetches
and locks, and load the Dag only when it is set.

execution_timeout is resolved from the same definition, so it is now sent
for a lang-SDK task only; #75387 has a Python task send its own to the
supervisor.

* Hold SQLite foreign keys off while the downgrade rebuilds

Dropping a column rebuilds the table on SQLite, unlike adding one, so the
downgrade copies task_instance rows while other tables still point at it.

* Pin the bulk-insert path and reword the is_stub comments

create_dagrun refreshes every row from its task right after the bulk write,
so the existing test passed even when insert_mapping dropped the key. Assert
the mapping itself instead, and record what a row keeps when it moves to a
new Dag version without being refreshed.

* Mark a Go or Java TriggerDagRun task a stub, as TypeScript does

* Inline the table name in the is_stub migration

---------

java-sdk/sdk/src/main/kotlin/org/apache/airflow/sdk/execution/Serde.kt

@@ -145,15 +145,16 @@ private fun serializeTask(
     // value never matches the tuple default it is compared against), so it
     // is unconditional here too. Java tasks have no template fields.
     data["template_fields"] = emptyList<Any?>()
-    // What marks a task whose arguments Airflow resolves per instance for a
-    // runtime outside Python, as `@task.stub` does on the Python side.
-    // `get_arg_bindings` reads nothing without it.
-    data["is_stub"] = true
     argBindings(taskId, def)?.let { data["_arg_bindings"] = it }
   }
+  // The Java runtime runs every task of a Java Dag, a trigger task included, so
+  // every one of them is a stub. is_stub marks a task whose arguments Airflow
+  // resolves per instance for a runtime outside Python, as `@task.stub` does on
+  // the Python side. `get_arg_bindings` reads nothing without it.
+  data["is_stub"] = true
   // Lets NotPreviouslySkippedDep re-skip a cleared downstream, as Python's SkipMixin does.
   if (def.decider != null) data["_can_skip_downstream"] = true
-  // A trigger task takes the Dag's queue too, because the Java runtime runs it.
+  // The worker picks a runtime by queue, so a trigger task takes the Dag's queue too.
   val config =
     if (dagQueue == null || "queue" in def.configValues) def.configValues else def.configValues + ("queue" to dagQueue)
   // Emit only config entries that differ from their schema default, mirroring

java-sdk/sdk/src/test/kotlin/org/apache/airflow/sdk/TriggerDagRunTest.kt

@@ -168,9 +168,10 @@ internal class TriggerDagRunTest {
     assertEquals("NOTSET", data["logical_date"])
     assertEquals(false, data["wait_for_completion"])
     assertEquals(false, data["skip_when_already_exists"])
-    // A Java task's markers belong to a task the Java runtime runs a body for.
+    // The Java runtime runs a trigger task too, so it carries the stub marker.
+    assertEquals(true, data["is_stub"])
+    // `language` marks a task the Java runtime runs a body for, which this is not.
     assertFalse("language" in data)
-    assertFalse("is_stub" in data)
     assertFalse("reset_dag_run" in data)
   }