feat(mobile): add OpenMindAI Speak read aloud

[smshagor-dev]

Summary

Adds local mobile read-aloud support as OpenMindAI Speak and carries the post-merge iOS CI signing fix onto a main-based branch.

OpenMindAI Speak

• assistant replies now expose a per-message read-aloud control
• active speech can be stopped from the same message action
• starting another reply stops the current utterance first
• sending, switching chats/models, regenerating, starting a new chat, or closing the screen stops speech cleanly
• Markdown is normalized for speech; fenced code blocks are omitted instead of reading syntax aloud
• Android uses the installed native TextToSpeech service and iOS uses native speech synthesis through the Flutter platform plugin
• no paid/cloud AI API is introduced

UI refactor

• extracts the message bubble and composer into focused widgets
• preserves camera/photo/file attachment, OpenMindAI Hear, generation stop/regenerate, model management and conversation search behavior

Native/CI

• adds Android TTS service discovery query required for Android 11+
• keeps Android minSdk 21 and iOS deployment target 15.6
• changes iOS CI from unsigned device build to simulator build so native plugins compile without Apple provisioning/signing

Validation target: Flutter analyze + tests + Android debug APK + iOS simulator build.