Add throughput

[matthewmturner]

• Based on output rows from IO
• Based on size of underlying data (go through the file groups from IO execs and read metadata on the files like file size etc). Would also be cool if we could use the parquet metadata functions to gather lower level details like row groups.