[DO NOT MERGE] Experimental Iceberg sharding runs

[ahmedabu98]

Experimenting with GroupIntoBatches' key parallelism when writing to Iceberg.

Main method is in IcebergShardingRuns.java.
Choose your own PROJECT, DATASET, and WAREHOUSE then configure the following options to run under different scenarios:

// ======== experiment with these numbers ===========
int numShards = 1;
long payloadSize = 1 << 10; // 1KB
int numIcebergPartitions = 0;
// ==================================================

Try to figure out what the best throughput-per-shard is that leads to large files while maintaining good performance.