[GLUTEN-11283][VL] Respect spark default zstd compression level

[wecharyu]

What changes are proposed in this pull request?

Spark default zstd compression leve is 3:
https://github.com/apache/parquet-java/blob/7ec32846663782846615e835bd06859c75b8f6b0/parquet-hadoop/src/main/java/org/apache/parquet/hadoop/codec/ZstandardCodec.java#L52

But velox default compression level is 1:
https://github.com/facebookincubator/velox/blob/a0bd1dd248c7def13dd17724e1569fd990515401/velox/dwio/parquet/writer/arrow/util/CompressionInternal.h#L76

So the output parquet size could be larger in gluten if user does not configure the parquet.compression.codec.zstd.level.

Fix #11283

How was this patch tested?

Test written file size locally.

[philo-he]

@wecharyu, I note Spark has the following config with zatd level set to 1 by default. Is it related? Do you know why Spark doesn't respect it?

https://github.com/apache/spark/blob/43f7936d7b3a4701e3d0fdb44663006cbe0db70b/core/src/main/scala/org/apache/spark/internal/config/package.scala#L2137

[wecharyu]

@philo-he The IO_COMPRESSION_ZSTD_LEVEL in package org.apache.spark.io is only used for internal data like eventlog writer, more discussion in SPARK-39743.

spark.hadoop.parquet.compression.codec.zstd.level can be used to set zstd compression level for table writer now.

[philo-he]

Thanks so much for the clarification. Looks good. cc @FelixYBW