feat: Add named parameter support for DuckDB read_parquet

[max-sixty]

Summary

Adds support for DuckDB's read_parquet optional boolean parameters to address issue #5548.

Users can now control DuckDB-specific behavior when reading Parquet files:

std.read_parquet 'data.parquet' union_by_name:true
std.read_parquet 'data.parquet' union_by_name:true binary_as_string:true

Changes

• Added 4 optional boolean parameters to read_parquet function:

  • binary_as_string (default: false) - Load binary columns as strings
  • file_row_number (default: false) - Include file_row_number column
  • hive_partitioning (default: null) - Interpret path as Hive partitioned
  • union_by_name (default: false) - Union columns by name instead of position

• Note: The filename parameter was intentionally excluded as it's deprecated since DuckDB v1.3.0 (automatically added as a virtual column)

Implementation Details

• Parameters are defined in std.prql with defaults
• DuckDB-specific SQL generation in std.sql.prql maps parameters to DuckDB's SQL named arguments
• Generic SQL implementation accepts parameters for signature compatibility but only uses source (consistent with PRQL's dialect handling)

Test Results

• ✅ All 608 core tests pass
• ✅ New tests verify correct SQL generation with named arguments
• ✅ Backward compatibility maintained (existing code continues to work)

Example SQL Output

read_parquet(
  'data.parquet',
  binary_as_string = false,
  file_row_number = false,
  hive_partitioning = NULL,
  union_by_name = true
)

Fixes #5548

🤖 Generated with Claude Code