Maybe actually build a `peek_next` function into the scanner

Currently we implement `peek_next` by cloning the scanner and advancing the clone. This honestly should be perfectly fine, since the clone is cheap and we don't need to call the function much anyway, but it's worth checking.

Start by locating the scanner and its current peek_next implementation. Compare the clone-and-advance behavior with the proposed direct scanner function, including any existing scanner tests. Done means the scanner provides peek_next without unnecessary cloning while preserving its current behavior.

Written by the indexing model from the issue text.