Add Parquet export path for research workflows

[erskingardner]

Problem

JSONL is the most universal exchange format, but many academic and data science workflows expect columnar formats for large-scale analysis. Pensieve currently does not provide a direct way to convert segment archives into Parquet.

Why this matters

A Parquet export can make large Nostr archive snapshots much cheaper to scan in DuckDB, Spark, Polars, Arrow, and similar tools. This lowers the barrier for researchers working with full-network datasets.

Suggested implementation

Add a Parquet export path after the JSONL exporter and shared segment reader exist.

Possible shape:

• Reuse common segment iteration code.
• Write event fields to Arrow/Parquet with a documented schema.
• Decide how to encode tags, likely as nested lists if supported cleanly, or as JSON strings if that keeps the first version simple.
• Support partitioning by segment, kind, or created_at date in a later iteration if needed.

Acceptance criteria

• Segment files can be exported to Parquet.
• The Parquet schema is documented.
• Export preserves all canonical Nostr event fields.
• Export handles tags deterministically.
• A small exported Parquet file can be read by DuckDB or an Arrow-compatible reader in tests or docs.
• just precommit passes before merging.