Lower row group size default to 512MB

[clairemcginty]

Lower row group size default to 512MB to relieve read-time memory pressure.

Note: Row group size is very much a tradeoff and there's no blanket one-size-fits-all best option. Each row group adds a bit of extra overhead to the file footer/row group dictionaries, however, for low-to-moderate cardinality datasets, the default compression algorithm (ZSTD) largely reduces any storage overhead.

Possibly we could go even smaller--Iceberg defaults to 128 MB row groups, for example. But I think this PR is a good starting point. In practice, without explicit configuration, Parquet writes over Dataflow seems to produce files in the realm of 200-300MB compressed (and 2-3x that size uncompressed)--thus, this change would result in a modest increase from 1 row group per file to 2.

(Another option is to log a warning on scio-smb writes that a smaller row group size is recommended, especially if the schema is narrow.)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 61.99%. Comparing base (ed9a1fb) to head (d1764c8).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #5993   +/-   ##
=======================================
  Coverage   61.98%   61.99%           
=======================================
  Files         323      323           
  Lines       11738    11743    +5     
  Branches      815      839   +24     
=======================================
+ Hits         7276     7280    +4     
- Misses       4462     4463    +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[clairemcginty]

moved this over to my fork of Scio: #5994