Derive the v4 content_stats type and stats field IDs #1127

Iceberg v4 replaces the per-column metric maps (lower_bounds, null_value_counts, ...) with a typed content_stats struct in tracked_file (spec Content Stats, apache/iceberg#14234). Its field IDs are derived from the table schema: each field's stats struct uses the IDs from 10_000 + 200 * field-id, plus a fixed offset per stat, and _last_updated_sequence_number and _row_id use the reserved bases 9000 and 9200. Which stats are present depends on the field type and the table's metrics mode.

Proposal: add a ContentStatsUtil with the stats ID math, the content_stats type a writer uses for a schema and MetricsConfig, and the projection a reader uses for a set of field IDs, matching Java's StatsUtil. The stats values and the bridge to the v1-v3 metric maps would be follow-ups.