fix(tests): resolve String addition compilation errors under Rust 1.96

[chandlerc]

Fix compilation failures under Rust 1.96.0+ caused by attempting to add &String to String (e.g. cannot add &String to String).

Resolved by converting string repetitions to borrowed string slices using .as_str() before addition. This ensures the entire integration test registry compiles successfully.

Assisted-by: Antigravity with Gemini

[rvben]

Thanks for the PR. I tried to reproduce this and could not: String + &String compiles on stable rustc 1.96.0 (the toolchain this repo pins via rust-toolchain.toml), on the current nightly 1.99.0 (2026-07-14), and the full test suite including the integration registry builds green on 1.96.0 in CI and locally. The impl Add<&str> for String plus deref coercion path this idiom relies on is unchanged upstream.

Could you share your exact rustc --version and the full error output? If your toolchain carries local patches this may be specific to it. I'd rather not take the churn without a rationale that reproduces on upstream Rust, but happy to revisit with a reproduction.

[chandlerc]

Doh, sorry about this. I had just updated rustc and so this was a plausible reason, but it wasn't accurate.

The real issue is that in a subsequent change I'm looking at adding use of a Rust crate that transitively pulls in the smartstring crate. That crate in turn includes another Add impl that breaks these +s.

I can fold the fix (this commit or an alternative) into the PR that adds the thing that uses them so it makes more sense.

A question: any preference about how we do this? There is the approach in this PR, but we could also more aggressively use format! to replace all the +s. Which approach would you prefer?

[rvben]

Thanks for the follow-up, that explains it. Folding it into the dependency PR sounds right, since there's nothing to merge standalone (upstream Rust compiles the current code fine).

On style: I'd keep the minimal .as_str() form you already have, limited to the sites the new Add impl actually breaks. format! reads fine in tests, but the non-test sites (md013, anchor_styles) benefit from + reusing the left-hand buffer, and I'd rather avoid a broader style migration motivated by a transitive dependency. Out of curiosity, which crate is coming in that pulls smartstring? That context will help when reviewing the follow-up PR.