nx_seed_sort.nx

nx_seed_sort.nx -- in-place sort of seed-pair arrays by (r, q). license_tier: INDEPENDENT_REDERIVE genealogy_id: international-research-sources/knuth-volume-3-insertion-sort G1.2b of NISHI_GENOMICS_SUBSTRATE_ROADMAP.md. Closes the composability gap between nx_align_match (G1.2) and nx_align_chain (G1.3): match outputs seed pairs in (outer-q, inner-r) order; chain DP requires pre-sorted-by-r input. This primitive bridges them so the seed-and-extend pipeline composes without test-time hand-sorting. Algorithm: classic insertion sort over paired (q, r) arrays, kept in lock-step. Sort key is (r_pos, q_pos) ascending lexicographic -- r is the primary axis (chain DP precondition) and q breaks ties deterministically so the leftmost-tie chain rule remains stable. Why insertion sort: O(n^2) reference impl is fine for short-read seed counts (n ~ 50-500); long-read or chromosome-scale needs a non-comparison sort (G1.5 nx_seed_sort_radix, 8-bit r-byte buckets). API: sort_seeds_by_r(seeds_q, seeds_r, n) -> i64 always 0; in-place nx_safety_envelope: (schema: nishi-library/seeds/safety-critical-standards.toml) intended_use: "Seed-pair sort bridging match -> chain; substrate primitive that removes the 'caller pre-sort' precondition from the chain DP" sil_target: SIL2 asil_target: QM dal_target: DAL C iec_62304_class: B evidence: [no_floating_point, deterministic, bit_equal_reproducible, insertion_sort_textbook, lex_tiebreak_stable, end_to_end_KAT_matcher_output_validates, license_tier_INDEPENDENT_REDERIVE] hazard_register: [bug-tape-paired-array-desync-on-swap,