[wtholliday]

Summary

Follow-up to #73.

• Count raw newlines inside string literals. The lexer only advanced the line counter for whitespace and comment newlines, so every token after a multi-line string reported the wrong line and a column measured from an earlier line. Raw newlines in a string now count, and the string token keeps the line of its opening quote.
• Replace the tok == Endl check with a pending_lines counter. Both the newline handler and the string scanner bump it, and the line advances when the next token is lexed. This removes the dependency on tok always holding the previously returned token. The parser's struct-literal backtrack saves and restores the counter with the rest of the lexer state.
• The doc comment on Lexer::loc claimed the column was computed on demand. It is set eagerly for every token.
• Add test_token_locations, pinning the Endl position at the newline on the line it ends, the deferred line advance across blank lines, the newline after a // comment, locations after a multi-line string, and where End lands after a trailing newline.

Before, for a stray } on line 5 after a two-line string:

str.lyte:4:1: Expected expression

After:

str.lyte:5:1: Expected expression

Test plan

• cargo test --workspace passes (unit tests plus all four golden runners)
• rustfmt --check clean on the touched files

🤖 Generated with Claude Code