[FEATURE]: make `opencode run` stream in token by token

[JimmyBenHur]

Feature hasn't been suggested before.

• [x] I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

In the interactive mode the responses are streamed token-by-token, but in the headless/non-interactive mode (opencode run) they don't seem to be. It would be very helpful, especially with local models, to see the responses in real time, instead of all at once after a few minutes.

[Catgroove]

This feature would be great since building ralph-like loops without it is pretty hard.