Count line-numbers during tokenization Before this PR, the `source_location` and `current_source_location` iterated (again) through input bytes separately from tokenization in order to count newline characters and determine the line number of some piece of a stylesheet. This PR makes this counting happen during tokenization instead, where we already have a pass looking at every bytes. <!-- Reviewable:start --> --- This change is [<img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/rust-cssparser/177) <!-- Reviewable:end -->