Unwrap AFM tool-calling scaffolding; retune two prompts for macOS 27 beta 4

The Foundation Models seed in macOS 27 beta 4 (26A5388g) answers a plain
`respond(to:)` as if it were driving a tool-calling harness, so free-form
replies now arrive wrapped: a bare JSON object, a fenced ```json block, a
`tool_call: {…}` literal, `[start_x]`/`[No tools needed]` markers, or a
persona preamble ahead of the answer. Every consumer takes the reply's
first non-empty line, so the wrapper landed verbatim in chat titles and
collapsed summaries — dev even carries a commit named
"Nucleic: {task:Update Website Hero Text}". Guided generation
(@Generable) still binds its schema and is untouched.

Add ModelOutput to NucleicProtocol — the one layer the Mac app, core and
the iOS client all link — and run it at each provider's text choke point,
before the existing line-oriented parsing. Plain replies pass through
byte-identical; an all-scaffolding reply yields nil, which is the "no
result" every caller already handles by falling back to its heuristic, so
no new failure path. DelegatedIntelligence unwraps again on receipt: the
mesh is mixed-version and the agent-CLI backend never unwraps.

Also retune the two prompts the seed broke, measured against the live
on-device model rather than by inspection:

- classifyTurn: the old wording let the agentically-tuned seed reason
  about what the agent should do NEXT, so it answered AWAITING for
  plainly finished turns — 12/17 on a labeled set (3 samples, majority
  vote) with 4 false AWAITING, which silently stops autoship. Reframing
  it as "a classifier, NOT an assistant" scores 17/17 with none. Guided
  generation was tried here and is worse (81% struct, 68% enum). Applied
  to both hand-duplicated twins.
- sessionKeywords: 9 of 9 runs returned a JSON/tool-call envelope, one
  inventing a fake Package.swift to "search". Anchoring extraction to
  terms already in the input returns 9 of 9 clean comma lists.

Tests pin the captured beta 4 shapes verbatim and the classifier clauses
that earned the accuracy, since they read like boilerplate.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

NucleicRemote/NucleicRemote/Models/PhoneIntelligenceExecutor.swift

@@ -69,8 +69,14 @@ enum PhoneIntelligenceExecutor {
             let deadline = request.deadlineSeconds ?? 60
             let text = await raceAgainstDeadline(seconds: deadline) {
                 await gate.run {
+                    // Strip the tool-calling envelope the macOS 27 / iOS 27 model wraps free-form
+                    // replies in before it goes on the wire — the host parses this text with the
+                    // same line-oriented heuristics as a local generation, so it must arrive
+                    // already unwrapped. A reply that was nothing but scaffolding yields nil, which
+                    // answers `error` below and drops the host straight to its heuristic.
+                    ModelOutput.usableText(
                         try? await LanguageModelSession(model: model, instructions: job.instructions)
-                        .respond(to: job.prompt).content
+                            .respond(to: job.prompt).content)
                 }
             }
             if let text, !text.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty {