Implemented batched inference for video analysis.

• updated the InferenceRunner classes to support batched inference
• added a detector_batch_size attribute to the project config.yaml (with default value 1)
• added batch_size and detector_batch_size parameters to video_inference_superanimal
• added batch_size and detector_batch_size parameters to analyze_videos (with default value None, which means that the batch sizes used are the ones defined in the config.yaml)
• added tests to ensure all frames are processed for top-down and bottom-up models