The events happens in the following order, assuming two media items:
1. First media item is fully decoded, record the last frame's pts
- Note frame processing is still ongoing for this media item
2. Renderer sends `onOutputFormatChanged()` to signal the second media item
3. **Block sending the frames of the second media item to the `VideoSink`**
4. Frame processing finishes on the first media item
5. The last frame of the first media item is released
6. **Reconfigure the `VideoSink` to apply new effects**
7. **Start sending the frames of the second media item to the `VideoSink`**
This CL implements the **events in bold**
PiperOrigin-RevId: 576098798
1fc34676d9