Feat/glue athena enhancements

[gettingstuffdone]

Description:
This PR completes the Athena fallback capability by automating Glue catalog setup for Parquet tables and improving SQL compatibility:

• Glue Provisioning for Parquet:
  • Extended GlueProvisioner to support the Parquet format with appropriate ParquetHiveSerDe, MapredParquetInputFormat, and
    MapredParquetOutputFormat.
  • Added serialization.format=1 parameter to ensure standard Hive compatibility.
• Athena SQL Translation:
  • Implemented _translate_to_athena to dynamically convert DuckDB-specific S3 scan functions (like read_parquet, iceberg_scan,
    read_csv, read_json) into qualified Glue table references (e.g., "database"."table_name") during the Athena execution path.
• Advanced Cost Estimation:
  • Added a sophisticated partition pruning detector (_has_partition_filter) using sqlglot to identify when a query can benefit from
    S3 prefix filtering.
  • Applied a partition_fraction (0.1) to cost estimates when valid partition filters are detected, providing more accurate "Dry Run"
    previews.
• Testing:
  • Added server/tests/test_cost_estimator_partitioning.py to verify partition pruning detection across basic, range, and complex
    IN/OR clauses.
  • Added Parquet-specific SerDe verification to server/tests/test_glue_provisioner.py.
  • Expanded server/tests/test_query_helpers.py with comprehensive SQL translation test cases.
• Dependency: This PR is based on feat/spend-tracking.

Note: Since these branches are sequential, this PR will show the changes from the refactor as well unless feat/server-refactor is merged first.

[endemics]

Thanks, much appreciated