Show byte size (MB, GB, etc.) of sources and model results

[ericpgreen2]

There are a few different sizes we could measure: compressed vs uncompressed, unloaded or loaded into DuckDB. To keep things directly comparable, we should probably use the data size that exists in DuckDB.

[hamilton]

We made an atttempt to do all of this back in february. It's much harder than it looks atm