feat: migrate on-device Qwen3 ASR to CoreML for background Flow dictation

Replace MLX GPU inference with CoreML bundles so transcription continues
while the host app is backgrounded. Adds model download and warm-up,
vendored Qwen3Speech, and updates onboarding, settings, and copy for the
~1.6 GB CoreML package (iOS 18+).

OSGKeyboardShared/Models/TranscriptionDelivery.swift

@@ -0,0 +1,17 @@
+// TranscriptionDelivery.swift
+// OSGKeyboard · Shared
+//
+// Host-app → keyboard handoff payload: final text plus an optional soft
+// warning when cloud polish failed but the raw transcript is still delivered.
+
+import Foundation
+
+public struct TranscriptionDelivery: Sendable, Equatable {
+    public let text: String
+    public let polishWarning: String?
+
+    public init(text: String, polishWarning: String? = nil) {
+        self.text = text
+        self.polishWarning = polishWarning
+    }
+}