fix(run): allow empty streamed model input

[hsusul]

Summary

This pull request fixes a streaming-only input restriction in Runner.run_streamed().

The non-streaming Runner paths pass an empty prepared item list to the configured model, but the streamed path raised RuntimeError("Prepared model input is empty") before model invocation. Removing that path-specific guard restores Runner parity and lets each configured model or provider decide whether it accepts empty input.

Test plan

• Added a public-interface regression test that streams a run from input=[], verifies the fake model receives the empty list, and verifies the run completes.
• Confirmed the test fails on pristine upstream/main with the expected RuntimeError.
• Ran the focused regression twice after the fix.
• Ran tests/test_agent_runner.py and tests/test_agent_runner_streamed.py: 212 passed.
• Ran .agents/skills/code-change-verification/scripts/run.sh: format, lint, typecheck, and the full test suite all passed.
• Ran git diff --check.

Issue number

Closes #3994

Checks

• I've added new tests, if relevant
• I've run .agents/skills/code-change-verification/scripts/run.sh
• I've confirmed all verification steps pass
• If using Codex, I've run /review before submitting this PR