Add --describe mode for summary statistics

[vmvarela]

Description

Load the data, print per-column statistics to stdout, exit. No query needed.

Examples

$ sql-pipe --describe sales.csv
Column     Type      NonNull  Null  Unique  Min         Max         Top3
─────────  ────────  ───────  ────  ──────  ──────────  ──────────  ──────────────────
id         INTEGER   12483    0     12483   1           12483       -
region     TEXT      12483    0     3       -           -           AMER(5200), EMEA(4100), APAC(3183)
amount     REAL      12400    83    8742    0.50        9999.99     -
date       DATE      12483    0     365     2024-01-01  2024-12-31  -

Acceptance Criteria

• --describe flag loads data and prints summary statistics
• Statistics include: column name, type, non-null count, null count, unique count, min, max
• For TEXT columns with low cardinality, show top 3 values with counts
• For numeric/date columns, show min and max values
• Output is formatted as an aligned table
• Works with file arguments and stdin
• Compatible with all input formats (CSV, TSV, JSON, NDJSON, XML)
• All existing tests pass
• New tests cover statistics computation and output formatting

Notes

• Implementation: ~300 lines
• Reuse type inference infrastructure
• Reuse table formatting code from pretty-print issue
• For large files (>100K rows), use approximate unique counts or skip
• Fits existing mode pattern (--columns, --validate, --sample, --describe)