nx_library_search.nx

nx_library_search.nx -- S-class search/retrieval over a LIBRARY of information (not one source). The Researcher recognizes an index page as a LIBRARY, harvests every doc it lists, then answers a SEARCH PROMPT by RANKING the whole corpus -- TF-IDF style, which exceeds naive substring search: * substring search can only say present/absent in one blob and cannot rank or pick the best doc * this ranks N docs by sum over query terms of tf(term,doc) * idf(term), where RARER terms (low document-frequency) weigh MORE -- so a doc that matches the distinctive query terms wins over a doc that merely mentions a common one. That is the retrieval quality real search engines use. Corpus is held as parallel arrays: ptrs[k] = (doc text *u8 stored as i64), lens[k] = its length. license_tier: ORIGINAL Pairs with nx_research_extract (re_count/re_has/re_find) + nx_library_cache.