[SPARK-37566][SQL] Fix repartitionByRange metrics

[cfmcgrady]

What changes were proposed in this pull request?

As the RangePartitioner does a sampling job in the input RDD, and the sampling job has the same execution context (execution id and metrics id) as a real job, then the number of output rows metrics of the RangePartitioner child will be increased by mistake.

spark.range(0, 10)
  .repartitionByRange(10, col("id"))
  .show()

Before this pr:

After this pr:

Does this PR introduce any user-facing change?

No, only bug fix.

How was this patch tested?

Added UT.