Support parquet as a loader format

Request: Support loading parquet as a source datatype in vega-loader

Envisioned solution:

{
  "data": [
    {
      "name": "my_data",
      "format": {"type": "parquet"},
      "url": "https://path/to/parquet.parquet",
    }
  ]
}

This could be accomplished via https://github.com/kylebarron/parquet-wasm.

I'm not sure, but would there be a way to push down the required columns into the data loader, ie. whenever a user in the frontend wants to change what they're visualizing, could you get those columns pushed down (so that you don't have to load/read/decode the entire parquet file?).

This would make vega super bleeding edge/powerful.