Module hypertable_pruning 

Hypertable chunk pruning for the SELECT planner.

Phase 0 of PRD #850 — activate the dormant partition-pruning primitive for hypertables. The machinery already exists (super::partition_pruning + the per-chunk time bounds tracked by HypertableRegistry) but nothing in the SELECT path consulted it, so every query against a hypertable scanned every chunk.

This module is the bridge: given a hypertable’s spec, its chunk set, and the SELECT WHERE clause, it returns only the chunks whose declared [start_ns, end_ns) interval can contain a row the temporal predicate admits. Chunks proven disjoint from the predicate window are dropped; everything else is kept.

Soundness contract — the pruner never drops a chunk that could hold a matching row. A row in chunk C carries a timestamp inside [C.start_ns, C.end_ns_exclusive); if any timestamp in that interval satisfies the predicate, C is kept. When the WHERE clause does not constrain the time column (or uses a shape the lowering can’t reason about), the pruner is conservative and keeps every chunk — exactly the Timescale / Postgres contract.

Functions

kept_scan_bounds

Smallest [lo, hi) nanosecond window that contains every kept chunk’s declared interval, or None when kept is empty.

prune_hypertable_chunks

Return the subset of chunks that may contain a row matching filter’s temporal predicate.