Fix memory leak in VecVideoRecorder by properly deleting references

[Copilot]

This PR fixes a memory leak in VecVideoRecorder where recorded_frames stayed in memory due to references held by the MoviePy clip object. The fix applies the same solution as implemented in Farama-Foundation/Gymnasium#1444.

Problem

When VecVideoRecorder stops recording and saves a video, the recorded_frames list would remain in memory because MoviePy's ImageSequenceClip maintains references to the frame data even after the video is written. This caused a memory leak during long training sessions with video recording enabled.

Solution

Added explicit del statements in the _stop_recording() method to properly clean up references:

def _stop_recording(self) -> None:
    """Stop current recording and saves the video."""
    assert self.recording, "_stop_recording was called, but no recording was started"

    if len(self.recorded_frames) == 0:  # pragma: no cover
        logger.warn("Ignored saving a video as there were zero frames to save.")
    else:
        from moviepy.video.io.ImageSequenceClip import ImageSequenceClip

        clip = ImageSequenceClip(self.recorded_frames, fps=self.frames_per_sec)
        clip.write_videofile(self.video_path)
        del clip  # ← Added this line

    del self.recorded_frames  # ← Added this line
    self.recorded_frames = []
    self.recording = False

The fix ensures that:

1. The MoviePy clip object is explicitly deleted after video writing
2. The old recorded_frames list is deleted before assigning a new empty list
3. Memory is properly released for garbage collection

Testing

• All existing tests continue to pass
• Added comprehensive memory cleanup verification tests
• Verified that recorded_frames are properly cleared after video saves
• Confirmed no regressions in video recording functionality

This change is backward compatible and only affects the internal memory management of the video recorder.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.