[Spark] Delete dataFile on iteration level in SparkParquetWritersFlatDataBenchmark

[yaooqinn]

This PR fixes a minor issue for a jmh test - SparkParquetWritersFlatDataBenchmark

The dataFile holds mocked outputs now cleaned at Level.Trial which will be executed after the set of benchmark iterations. Then, it cause errors like below.

# Warmup Iteration   1: 7.251 s/op
# Warmup Iteration   2: <failure>

org.apache.iceberg.exceptions.AlreadyExistsException: File already exists: /Users/kentyao/iceberg/spark/v3.2/spark/build/tmp/jmh/parquet-flat-data-benchmark3385433263675547096.parquet
	at org.apache.iceberg.Files$LocalOutputFile.create(Files.java:58)
	at org.apache.iceberg.parquet.ParquetIO$ParquetOutputFile.create(ParquetIO.java:148)

In this PR, I set the TearDown level to Iteration to make the dataFile be deleted after each iteration.

[rdblue]

@nastra or @RussellSpitzer, can you review this one? Thank you!

[nastra]

LGTM

[nastra]

@rdblue or @RussellSpitzer when you merge this one, please also merge #3910 as it fixes the same issue for Spark 3.2 benchmarks

[rdblue]

Thanks, @yaooqinn!