nx_outlink_harvest.nx

nx_outlink_harvest.nx — shared outlink -> link-graph-edge harvester (P1 PageRank feed). LIB (no main). Extracted (by copy) from nx_web_crawl_step.wc_harvest so BOTH the live crawler AND nx_cc_ingest emit the same row shape: out:<ci_hash(page_url)> = packed i64 ci_hash(target_url). Node identity = ci_hash(url) == dss_urlcid(url) at serve time, so an outlink resolves to a graph node with zero extra index; serve-time authority = content_cid -> url:<cid> -> ci_hash(url) -> pr:<cid>. ABSOLUTE cross-site links (the authority- bearing ones) produce byte-identical cids to the crawler. Three intentional improvements over the crawler's inline copy (all to backport when the crawler adopts this lib): (1) root-relative "/path" resolves HOST-based (scheme://host + path) not base+path — the crawler's concatenation makes ".../Trust_law" + "/about" a phantom node; (2) edge capture is DECOUPLED from the frontier cap (the crawler stops recording edges once discbox hits WC_MAXDISC — a graph must not depend on frontier state); (3) one REUSED url buffer instead of an mmap per outlink (bounds address-space growth for CC-scale ingests of thousands of pages). license_tier: ORIGINAL