pub fn tokenize(text: &str) -> Vec<String>
Lowercase terms split at camelCase boundaries and non-alphanumerics, without stop words.