fix: propagate send failures, guard image conversion, cleanup speech mic (#130)

Seven correctness bugs in the session composer:

- sessions.ts sendMessage: await the prompt submission and rethrow on
  failure instead of a fire-and-forget .catch(), so handleSend's existing
  restore-draft-and-alert catch actually runs.
- pasteFromClipboard: route pasted images through toJpeg() so they get
  the same resize/compress treatment as picked/captured photos.
- pickFromLibrary/pickFromCamera: wrap toJpeg() in try/catch (and switch
  to Promise.allSettled for the multi-select batch) so one bad asset
  doesn't silently drop the whole batch; surface a new imageFailed alert.
- pickFromLibrary: cap selection at 10 images.
- useSpeech: abort the native recognition session on unmount so the mic
  doesn't stay hot after leaving the screen.
- Surface useSpeech's error via Alert, keyed on the error value so it
  fires once per distinct error.
- Undo on the revert banner now also clears the composer, since it was
  prefilled by the edit flow and could otherwise be sent as a duplicate.


Claude-Session: https://claude.ai/code/session_01T12AhSnQVrSxNnvwfCx2z6

Co-authored-by: engineer <engineer@macbookpro.lan>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

src/lib/i18n/en.json

@@ -118,7 +118,11 @@
       "replyFailedTitle": "Reply Failed",
       "replyFailedMessage": "Could not send your response. Please try again.",
       "rejectFailedTitle": "Reject Failed",
-      "rejectFailedMessage": "Could not send your response. Please try again."
+      "rejectFailedMessage": "Could not send your response. Please try again.",
+      "imageFailedTitle": "Image not attached",
+      "imageFailedMessage": "One or more images could not be processed. Please try a different photo.",
+      "speechErrorTitle": "Voice input failed",
+      "speechErrorMessage": "Could not use voice input. Check your microphone permission and try again."
     }
   },
   "connection": {

src/lib/i18n/zh-Hans.json

@@ -118,7 +118,11 @@
       "replyFailedTitle": "回复失败",
       "replyFailedMessage": "无法发送您的回复。请重试。",
       "rejectFailedTitle": "拒绝失败",
-      "rejectFailedMessage": "无法发送您的回复。请重试。"
+      "rejectFailedMessage": "无法发送您的回复。请重试。",
+      "imageFailedTitle": "图片未添加",
+      "imageFailedMessage": "一张或多张图片无法处理。请尝试其他照片。",
+      "speechErrorTitle": "语音输入失败",
+      "speechErrorMessage": "无法使用语音输入。请检查麦克风权限后重试。"
     }
   },
   "connection": {

src/lib/speech.ts

@@ -1,4 +1,4 @@
-import { useState, useCallback, useRef } from "react"
+import { useState, useCallback, useRef, useEffect } from "react"
 import { ExpoSpeechRecognitionModule, useSpeechRecognitionEvent } from "expo-speech-recognition"
 
 interface SpeechState {
@@ -76,5 +76,13 @@ export function useSpeech(onResult: (text: string) => void): SpeechState & Speec
     setTranscript("")
   }, [])
 
+  // Stop the native recognition session when the screen unmounts — otherwise
+  // the mic stays hot in the background. abort() is a no-op when not listening.
+  useEffect(() => {
+    return () => {
+      ExpoSpeechRecognitionModule.abort()
+    }
+  }, [])
+
   return { listening, transcript, error, start, stop, cancel }
 }