Add maximal-munch regression tests for Wave operators

## Why

`front/lexer/src/scan.rs` recognizes many overlapping operators such as `+`/`++`/`+=`, `?`/`??`/`?:`, and `:`/`::`. There are currently no lexer unit tests protecting the longest-token rule.

## Scope

- Add a table-driven lexer test for every single- and multi-character operator.
- Include adjacent operator sequences and an EOF immediately after an operator.
- Assert that multi-character operators are emitted as one token and not split into prefixes.
- Include punctuation used by imports, generic types, arrays, and function return arrows.
- Do not alter precedence or parser semantics.

## Likely files

- `front/lexer/src/scan.rs`
- `front/lexer/src/token.rs`
- A new lexer integration test

## Completion criteria

- [ ] All operator variants are covered.
- [ ] Ambiguous-prefix cases are explicit regressions.
- [ ] Token lexemes and line numbers are asserted.
- [ ] `cargo test -p lexer --locked` passes.

Read front/lexer/src/scan.rs and front/lexer/src/token.rs to map the operator tokens, then inspect the existing lexer tests or add the proposed integration test. Run cargo test -p lexer --locked; done means every operator variant, adjacent sequence, EOF case, ambiguous prefix, lexeme, and line number is asserted without changing parser semantics.

Written by the indexing model from the issue text.