Add dataversion utility for extracting version info from data files

[ccdavis]

Add new data_version module and dataversion CLI binary to extract version metadata from IPUMS data files:

• For parquet: reads all key-value metadata from file_metadata, counts variables, ignores samples/datasets, outputs everything else as version info
• For fixed-width: reads system variables (record type '#') from layout, parses first line of compressed data file to extract values

Most IPUMS fixed-width will only have 'ZZZZ...' values for version variables since Workspaces requires these don't vary unless data changes -- this is for supporting the data checksum == data version behavior. Mainly you'd use it on the parquet data.

It's meant to support reporting on what's deployed to servers, but could be used by various deployment or data reproducibility tasks.
Right now the data_version module repeats some functionality found in rsvp relating to parquet files and path conventions and fixed width stuff. I plan to eventually pull in the fixed-width support and multi file parquet reading into cimdea and then that extra code can go away.

Features:

• Auto-detects file format from path (.parquet/.dat.gz)
• Recognizes parquet datasets by convention (parent dir named "parquet")
• Handles both single parquet files and partitioned parquet datasets
• Dynamic metadata extraction - no hardcoded field names
• Output formats: human-readable text or JSON
• Reuses existing DatasetLayout parser for fixed-width layouts

Usage:
dataversion /path/to/parquet/us2015b dataversion /path/to/us2015b_usa.dat.gz dataversion --format json /path/to/data

Partially 🤖 Generated with Claude Code