Default to snappy compression for geoparquet

[gadomski]

Requested in #74 (comment), and Python's parquet defaults to snappy: https://arrow.apache.org/docs/python/generated/pyarrow.parquet.write_table.html#pyarrow-parquet-write-table

[hrodmn]

I remember seeing a Github thread about stac-geoparquet compression and the recommendation was to go with zstd. I can't find it right now but I will keep looking!