AWS S3 Benchmarks for Parquet

[stanbrub]

Deephaven now has the ability to read parquet stored in the ASW S3 cloud storage. Benchmarking this requires use of MinIO running on the Benchmark server to take WAN latency out of the equation.

• Add MinIO to standard docker-compose.yml for local testing
• Write benchmarks for read parquet (like the ParquetRead MultiCol tests already existing)
• Requires Per Benchmark Docker Service Selection #314
• A round-trip has been done to proof the concept, but there are questions:
  • Will we do partial reads like we do with 'view'?
  • Will pushdown predicates work with s3?
  • Are we testing with any special repo settings? Like compression?
  • How much memory is required for MinIO for adequate testing?