[Optimize]: coalesce streamed text chunks before Tauri emit to reduce IPC and peer fanout pressure

[YodonTan]

Problem / opportunity

During streaming AI responses, the agent runtime emits one TextChunk / ThinkingChunk event per provider chunk. The desktop event loop forwards every chunk to the WebView as a separate Tauri IPC message (JSON serialization, WebView2 boundary crossing, JS parse + dispatch) and, when peer devices are attached, as a separate end-to-end encrypted relay message. Message volume scales with chunk rate, so fast streams create avoidable IPC and network pressure.

Proposed behavior

Coalesce streamed text/thinking chunks before the transport emit in the desktop event loop:

• Merge chunks sharing the same (session, turn, round, attempt, contentType) within a ~50ms time window; text is concatenated and the thinking isEnd flag is OR-ed.
• Non-chunk events (completion, error, cancellation, tool events) flush pending merged text first and then pass through, preserving per-stream ordering.
• Critical / High priority events pass through immediately.
• Content stays equivalent (the frontend already accumulates the same text), ordering is preserved, and internal event subscribers are unaffected.

Product surface

Desktop app / Cross-cutting

Details, examples, or constraints

• Reduces Tauri IPC messages by roughly 3-10x during fast streams and proportionally reduces remote peer fanout (each event currently costs serialize + encrypt + base64 + relay + decrypt/parse on the mobile side).
• Adds at most ~50ms to text delivery; the frontend typewriter and the existing 32ms EventBatcher absorb this.
• Only the desktop transport path changes: EventQueue, event contracts, ACP/TUI/CLI consumers, and frontend code are untouched.