Accelerate the coalescing parquet reader when reading files from multiple partitioned folders

[tgravescs]

fixes #1200

Accelerate the scan speed for coalescing parquet reader when reading files from multiple partitioned folders.

Previously whenever we hit a file that was in a different partition we split the batch so we could easily add the partition values. This results in us having to do a lot of batches when there are a lot of partitioned files, which is not great for performance.
To fix that we can change it to combine files with different partitioning by keeping track of the partition values and which rows those values need to be applied to. Then after we read the files we need to add those columns that are built based off the partition values and row counts. This works because we read the files in the same order as when we construct what goes into each batch.

in this PR I added the tracking of partition values and the corresponding number of rows, then after we read the files into the columnar batch we add all the partition values. The partition values are constructed by doing 1 partition column at a time. For each column it generates the individual partition value columns for the number of rows necessary, then it concatenates all of those together, then it moves to the next partition column. ie meaning if you have paths with multiple partitions ../key1=2/key2=foo/ , it does key1=X for all values of X first then it does key2=Y afterwards.

[tgravescs]

build

[github-actions]

👎 Promotion blocked, new vulnerability found

Vulnerability report

Component	Vulnerability	Description	Severity
Guava: Google Core Libraries for Java	CVE-2020-8908	A temp directory creation vulnerability exist in Guava versions prior to 30.0 allowing an attacker with access to the machine to potentially access data in a temporary directory created by the Guava com.google.common.io.Files.createTempDir(). The permissions granted to the directory created default to the standard unix-like /tmp ones, leaving the files open. We recommend updating Guava to version 30.0 or later, or update to Java 7 or later, or to explicitly change the permissions after the creation of the directory if neither are possible.	LOW

[jlowe]

The original code would always close fileBatch, but this now leaks it if something throws before we get to addGpuColumnVectorsToBatch (e.g.: buildPartitionColumns).

[tgravescs]

I think I'm handling this now, let me know if you see anything I missed.

[pxLi]

build

[pxLi]

build

[tgravescs]

thanks Jason, updated.

[tgravescs]

build

[jlowe]

Just a small nit for what appears to be an unnecessary null check but otherwise looks good to me.

[tgravescs]

build