Add streaming output for live evaluation progress

[aviralgarg05]

Summary

Implement streaming output support for long-running evaluations.

Tasks

• Add --stream flag to CLI that prints results as scenarios complete.
• Use rich/click progress bars for visual feedback.
• Show running metrics (pass/fail count, current scenario).
• Support both TTY and non-TTY output modes.

Acceptance criteria

• agentunit eval scenarios/ --stream shows live progress.
• Works correctly when piped to file.
• No buffering issues on slow evaluations.