.why: no KG deep copy, indexed subgoal proofs, release storage guard #134

execute_and_get_context clones all input tuples per call (storage_engine/mod.rs:769). Backward chaining proves each subgoal by linear scan (backward_chaining.rs:498, unification.rs:72). why_query holds the storage read guard throughout. Pass Arc maps, build per-(relation, bound columns) indexes once per call, and drop the guard after taking the snapshot. The gateway runs .why for every finding on every turn, so this is on the hot path.

Roadmap checkpoint — 2026-10-03

S06 (#166) owns the immediate recursive-read deadlock in R1. R04 here still owns snapshot/Arc relation reuse, indexed bound-subgoal proof construction and releasing storage before expensive proof work. Verify tuple equality before/after, concurrent writes, large result paging and gateway changed-finding proof cost; pass #163. S06 alone does not complete this issue.