Feature Request: Typing/Thinking Indicator Support

Problem

When using slower models (e.g., Kimi K2.5) or complex reasoning tasks, users don't know if the agent is processing or stuck. This creates a poor UX.

Proposed Solution

Add support for sending intermediate "thinking/processing" indicators:

1. iMessage: Use typing indicator (if protocol supports)
2. Feishu/Lark: Support typing message type or emoji reactions
3. Other channels: Send intermediate "Processing..." message before final response

Use Cases

• Long-running reasoning tasks
• Multi-step tool executions
• Slow model inference (Kimi, GPT-4, etc.)

Technical Approach Options

Option A (Safe): Send "Processing..." text message immediately, then edit/update with final response
Option B (Native): Use platform-specific typing indicators where available

Priority

Medium - Nice to have for better UX, but not blocking

/cc @maintainers