Feature request: show live tokens/sec (output speed) in TUI and Desktop app #50000

Description

When the model is generating a response, there's no visible indication of output speed (tokens per second). The UI currently shows token usage and cost per message/session, but no live tok/s metric while streaming.

Use case

• Comparing provider/model latency and throughput side by side
• Quickly spotting slow providers, rate limiting, or degraded performance
• Diagnosing whether long waits are network/provider related or model-side

Suggested implementation

• Show output tok/s next to the streaming indicator while a response is generating
• Optionally include tok/s in the per-message token usage details after completion

Tested on the v2 Desktop app; the same gap exists in the TUI.