[SPARK-46585][CORE] Directly constructed metricPeaks as an immutable.ArraySeq instead of use mutable.ArraySeq.toSeq in Executor

[LuciferYang]

What changes were proposed in this pull request?

The following code exists in branch-3.5:

spark/core/src/main/scala/org/apache/spark/executor/Executor.scala

Lines 744 to 749 in fb90ade

	// Here and below, put task metric peaks in a WrappedArray to expose them as a Seq
	// without requiring a copy.
	val metricPeaks = WrappedArray.make(metricsPoller.getTaskMetricPeaks(taskId))
	val reason = TaskKilled(t.reason, accUpdates, accums, metricPeaks.toSeq)
	plugins.foreach(_.onTaskFailed(reason))
	execBackend.statusUpdate(taskId, TaskState.KILLED, ser.serialize(reason))

From the code comments, when using Scala 2.12, metricsPoller.getTaskMetricPeaks(taskId) is wrapped as WrappedArray to avoid collection copying, and the subsequent metricPeaks.toSeq is a redundant collection conversion for Scala 2.12.

However, for Scala 2.13, if metricsPoller.getTaskMetricPeaks(taskId) is still wrapped as WrappedArray/mutable.ArraySeq, it is impossible to avoid collection conversion, because the subsequent metricPeaks.toSeq will trigger a collection copy when using Scala 2.13.

So this pr changes the process to directly wrap metricPeaks as immutable.ArraySeq to ensure the same effect as when using Scala 2.12, which is also more in line with the original comment description.

Why are the changes needed?

Avoid unnecessary collection copying when using Scala 2.13.

Does this PR introduce any user-facing change?

No

How was this patch tested?

Pass GitHub Actions

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

No

[dongjoon-hyun]

+1, LGTM.

[LuciferYang]

Merged into master for Spark 4.0. Thanks @dongjoon-hyun ~