Added comprehensive unit tests that verify:
- Text-only modality filters out audio outputs
- Audio-only modality filters out text but preserves usage info
- Multiple modalities includes all requested outputs
- No modalities parameter includes all outputs
These tests use mocks to run quickly without requiring full model loading.
Co-authored-by: hsliuustc0106 <222337142+hsliuustc0106@users.noreply.github.com>