[Parquet] Add tests for IO/CPU access in parquet reader

parquet/Cargo.toml

@@ -78,6 +78,7 @@ base64 = { version = "0.22", default-features = false, features = ["std"] }
 criterion = { version = "0.5", default-features = false, features = ["async_futures"]  }
 snap = { version = "1.0", default-features = false }
 tempfile = { version = "3.0", default-features = false }
+insta = "1.43.1"
 brotli = { version = "8.0", default-features = false, features = ["std"] }
 flate2 = { version = "1.0", default-features = false, features = ["rust_backend"] }
 lz4_flex = { version = "0.11", default-features = false, features = ["std", "frame"] }

parquet/src/file/reader.rs

@@ -48,11 +48,12 @@ pub trait Length {
 /// Generates [`Read`]ers to read chunks of a Parquet data source.
 ///
 /// The Parquet reader uses [`ChunkReader`] to access Parquet data, allowing
-/// multiple decoders to read concurrently from different locations in the same file.
+/// multiple decoders to read concurrently from different locations in the same
+/// file.
 ///
-/// The trait provides:
-/// * random access (via [`Self::get_bytes`])
-/// * sequential (via [`Self::get_read`])
+/// The trait functions both as a reader and a factory for readers.
+/// * random access via [`Self::get_bytes`]
+/// * sequential access via the reader returned via factory method [`Self::get_read`]
 ///
 /// # Provided Implementations
 /// * [`File`] for reading from local file system