feat: add standalone shuffle benchmark binary for profiling

Add a `shuffle_bench` binary that benchmarks shuffle write and read
performance independently from Spark, making it easy to profile with
tools like `cargo flamegraph`, `perf`, or `instruments`.

Supports reading Parquet files (e.g. TPC-H/TPC-DS) or generating
synthetic data with configurable schema. Covers different scenarios
including compression codecs, partition counts, partitioning schemes,
and memory-constrained spilling.

Author: andygrove

native/Cargo.lock

@@ -72,6 +72,7 @@ url = { workspace = true }
 aws-config = { workspace = true }
 aws-credential-types = { workspace = true }
 parking_lot = "0.12.5"
+clap = { version = "4", features = ["derive"] }
 datafusion-comet-objectstore-hdfs = { path = "../hdfs", optional = true, default-features = false, features = ["hdfs"] }
 reqwest = { version = "0.12", default-features = false, features = ["rustls-tls-native-roots", "http2"] }
 object_store_opendal = {version = "0.55.0", optional = true}
@@ -113,6 +114,10 @@ name = "comet"
 # "rlib" is for benchmarking with criterion.
 crate-type = ["cdylib", "rlib"]
 
+[[bin]]
+name = "shuffle_bench"
+path = "src/bin/shuffle_bench.rs"
+
 [[bench]]
 name = "parquet_read"
 harness = false