pub fn tokenize(text: &str) -> Vec<String>Expand description
Split text into lowercased alphanumeric words (length >= 2).
find_text semantics are substring-based, so query tokens are treated the
same way here and candidates are later verified with the exact predicate.