perf(persona): drop per-turn format!()+lowercase — use canonical RoleId::as_str() (#195 slice 3)

[joelteply]

Summary

Slice 3 of task #195 (inference latency perfection campaign). Drop the per-turn format!("{:?}", ctx.role).to_lowercase() from serve_persona_loop_inner and read the canonical RoleId::as_str() directly. No new field on PersonaContext, no cache plumbing — as_str()'s &'static str IS the cache.

What ships (one-line behavior change + one test)

// service_loop.rs, in RespondInput construction:
// was: specialty: format!("{:?}", ctx.role).to_lowercase()
specialty: ctx.role.as_str().to_string(),

Per-turn cost drops from format!() (variadic macro + Debug formatter + heap alloc) + .to_lowercase() (second alloc + Unicode lowercase walk) to a single String::from of a static &'static str.

Why this shape (not a cache)

Initial draft of slice 3 added a PersonaContext.specialty: Arc<str> cache populated by a build_persona_specialty(role) helper. Three adversarial reviewers correctly flagged:

1. Caching the wrong source. RoleId::as_str() is the documented, intentionally-stable specialty derivation (kebab-case identifier, pinned by an explicit match in role_template.rs). Caching the Debug+lowercase chain instead couples prompt assembly to the derived-Debug format — a future variant rename or wrapper struct could silently break the specialty with no test failure.
2. The cache test was a tautology. The helper was Arc::from(format!("{:?}", role).to_lowercase()) and the test compared its output to format!("{:?}", role).to_lowercase() — comparing the helper to its own implementation.

The revised slice picks the right source and the right contract: use role.as_str() directly. No cache because the static-str pointer IS already pre-computed at compile time. The new test is non-circular: it compares two independently-derived strings (role.as_str() from role_template.rs::match, vs the pre-slice-3 derived-Debug chain).

Test

role_as_str_preserves_pre_slice3_specialty_format_for_each_role:

• Pins that for EVERY RoleId variant (Helper, Coder, Sentinel, Custom), role.as_str() produces byte-identical output to the pre-slice-3 format!("{:?}", role).to_lowercase(). Non-circular.
• Compile-time exhaustiveness: a closure-typed |role: RoleId| match role { ... } with NO wildcard arm forces a new RoleId variant to be added here AND to the variants array. (Initial draft's "cosmetic exhaustiveness" — matching on a literal RoleId::Helper — was fixed.)

Out of scope

• No display_name change. String::clone of a fixed-per-session name has the same per-turn cost cache-or-not; the smell is RespondInput's String-typed field, not the source. Lift when RespondInput's type changes (Build(deps-dev): Bump @typescript-eslint/eslint-plugin from 8.29.1 to 8.46.2 #149-adjacent).
• No prompt-template change. Substrate keeps the existing specialty-string contract verbatim; the migration is provably behavior-preserving.

Doctrine

• [[init-once-handle-then-lease-zero-copy-refs]] — canonical "init-once" here is the compile-time match in RoleId::as_str(). Per-turn cost is a &'static str deref + one String::from (memcpy of 6-9 static bytes). Doesn't get cheaper without changing RespondInput's type signature.
• [[observability-is-half-the-architecture]] — slice 1's respond_latency aggregate will tell us if this + cumulative slice-2/3+ wins are moving the needle on real persona turns.
• [[no-fallbacks-ever]] — no graceful-degradation; the behavior-preservation test breaks loudly if either side drifts.

card: 0d780926
parent task: #195
slice 1: #1532 (merged)
slice 2: #1533 (merged)