99Build a lightweight client-side search index for 10,000 records

Filtering 10,000 records with includes on every keystroke is slow and gives poor results. Build a small search engine:

• tokenize(text): lower-case, remove accents (Café becomes cafe), and split on anything that is not a letter or digit. "Café-Crème, 2024!" gives ["cafe", "creme", "2024"].
• createSearchIndex(records, { fields, idField = "id", weights = {} }) builds an inverted index (token → records containing it) and returns { search(query, { limit = 10 }) }.

search rules:

• Every query token must match (AND).
• The last token also matches as a prefix, so results appear while the user is still typing ("reac" finds React).
• Score = the sum of field weights (default 1) for every occurrence of a matching token; a prefix match counts half.
• Return the matching records sorted by score (higher first), ties in original order, at most limit.
• An empty query returns [].

It must build an index of 10,000 records and answer 1,000 queries in well under a second.