Add Parquet format support for output (and input)

[dlevy-msft-sql]

Summary

Add support for reading and writing Parquet files in sqlcmd, similar to how CSV and other formats are supported.

Motivation

Parquet is a columnar storage format widely used in data engineering and analytics pipelines. Supporting Parquet would enable:

• Efficient data export: Column-oriented compression for large query results
• Interoperability: Direct integration with tools like Spark, Pandas, DuckDB, Polars, and cloud data lakes
• Type preservation: Parquet preserves data types better than CSV (dates, decimals, nulls)
• Performance: Faster reads for analytical workloads due to columnar format

Proposed Features

Output (-o parquet\ or similar)

• Export query results to Parquet files
• Support for common compression codecs (snappy, gzip, zstd)
• Proper SQL Server to Parquet type mapping

Input (optional, lower priority)

• Read Parquet files as input for bulk insert operations
• Could complement -i\ flag for file input

Implementation Considerations

• Go Parquet libraries: parquet-go, apache/arrow-go
• Type mapping from SQL Server types to Parquet logical types
• Memory management for large result sets (streaming vs buffered)
• Schema inference vs explicit schema definition

Related

• Similar to existing CSV output functionality
• Would complement the data export/import story for go-sqlcmd

Tasks

• Research Go Parquet libraries and choose one
• Design CLI flags (-f parquet, --parquet-compression, etc.)
• Implement SQL Server → Parquet type mapping
• Implement streaming Parquet writer for query results
• Add tests
• Document usage
• (Optional) Implement Parquet input for bulk operations