Move the Rust toolchain pin from 1.97.1 to 1.98.0 (#120)

[kurok]

Closes #120.

Why the pin can move

The pin was a holding action against a measured 15–30% slowdown under rustc 1.98.0. The cause turned out not to be the compiler's codegen: two byte-at-a-time loops in mailparse ran at half speed when the linker placed them across a 64-byte boundary on the runners' Zen CPUs, and a new rustc — like a version bump, #204 — moved them. Their x86-64 instruction streams were byte-identical under both compilers (find_from_u8_line_prefix: 88 = 88; decode_base64: 147 = 147).

With both loops replaced (#213, #214), the toolchain A/B on the CPU that showed the worst of it — the EPYC 9V74, +96% this morning, +22% after #213 alone — now reads 1.98.0 within ±0.5% of 1.97.1 on every parse path (run 33092063031, noise floor 4.0%, worst +2.3% on a metadata batch). Locally on an M4, same source under both compilers: worst +0.9%.

Why it stays a pin rather than floating to stable

The benchmark gate builds a revision and its base with the same toolchain, so a compiler change is the one regression it cannot see — both sides move together. A floating stable would put an unmeasured compiler into every published wheel. Keeping a pin makes the compiler a change like any other: proposed, measured with toolchain-ab.yml, reviewed. rust-toolchain.toml now says this and describes the bump procedure; the toolchain-ab.yml baseline default follows the pin.

What changed

• rust-toolchain.toml: 1.97.1 → 1.98.0; comment rewritten (the old one quoted the ~26% figure Investigate the ~26% slowdown under rustc 1.98.0, then unpin the toolchain #120 later showed to be a measurement artifact, and a 7.0x floor the gate no longer has).
• test.yml (3 dtolnay inputs + maturin-action), publish.yml (4 maturin-action inputs): 1.98.0, because components must be installed for the toolchain cargo actually uses (precedence note in the toolchain file).
• toolchain-ab.yml: baseline default and history comment.
• CONTRIBUTING prerequisites + Performance intro; CHANGELOG.

Verified locally under 1.98.0 before pushing: cargo fmt --check clean (blocking in CI), clippy unchanged from master, vendored crate's tests pass, 803 tests pass on a 1.98.0-built wheel.