fn classify_token(source: &str, offset: usize) -> String
Classify the token found at a byte offset in the source.