[iOS] Controlled single-line TextInput duplicates dictated text (New Architecture) #59406

Description

On iOS 16+ with the New Architecture, ending dictation (tapping the keyboard mic key again) in a single-line controlled TextInput inserts the dictated text twice. Uncontrolled inputs and the old architecture are not affected.

RCTTextInputComponentView only falls back to a plain-text comparison during dictation when textInputMode.primaryLanguage == "dictation", which iOS 16+ no longer reports. The old architecture also checks dictationRecognizing (#37188, fixing #36045); Fabric doesn't. Adding that check fixes it (0/5 duplicated vs 5/5 without).

Steps to reproduce

1. Run the reproducer on a physical iPhone or iPad.
2. Tap the "Controlled" field, tap the keyboard mic key and say a sentence.
3. Tap the mic key again: the sentence appears twice.

React Native Version

0.88.0

Affected Platforms

Runtime - iOS, Build - MacOS

Areas

Fabric - The New Renderer

Output of npx @react-native-community/cli info

System:
  OS: macOS 15.6
  CPU: (8) arm64 Apple M1
Binaries:
  Node:
    version: 22.14.0
  npm:
    version: 10.9.2
  Watchman:
    version: 2025.03.10.00
Managers:
  CocoaPods:
    version: 1.16.2
IDEs:
  Xcode:
    version: 16.2/16C5032a
npmPackages:
  "@react-native-community/cli":
    installed: 20.2.0
  react:
    installed: 19.3.0
  react-native:
    installed: 0.88.0
iOS:
  hermesEnabled: true
  newArchEnabled: true

Stacktrace or Logs

N/A (no crash)

MANDATORY Reproducer

https://github.com/paulnewby/rn-dictation-reproducer

Screenshots and Videos

Device: iPad (10th generation), iPadOS 26.