pub fn estimate_tokens(content: &str) -> usize
Estimates token usage using an average of four UTF-8 bytes per token.