Refactor model configuration with a model registry

[xiu-cs]

This pull request updates the model loading and configuration system for FMPose3D and Animal3D workflows, moving from direct file path imports to a model registry approach. This makes it easier to specify models by name, improves reproducibility, and standardizes script arguments. It also updates pretrained model references and cleans up related scripts.

Model loading and registry improvements:

• Replaced direct imports of model classes with a registry system using get_model and --model_type for both human and animal 3D pose estimation; updated code in main_animal3d.py, vis_animals.py, and FMPose3D_main.py to use this registry. [1] [2] [3]
• Registered animal 3D model (Model) with the registry via @register_model("fmpose3d_animals") and made it inherit from BaseModel.
• Updated model argument parsing to add --model_type (defaulting to fmpose3d_animals) and clarified the use of --model_path as an override.
• Ensured all model subpackages (including animal models) are imported so their registry decorators execute.

Script and configuration updates:

• Updated training, testing, and visualization shell scripts (train_animal3d.sh, test_animal3d.sh, vis_animals.sh, FMPose3D_train.sh, FMPose3D_test.sh) to use --model_type instead of --model_path, and set new default paths for pretrained weights. [1] [2] [3] [4] [5] [6] [7] [8] [9]
• Changed the Animal3D pretrained model download link in the documentation and updated instructions accordingly.

Minor improvements and cleanup:

• Removed legacy code for copying local model files in experiment checkpointing.
• Minor code and comment cleanups in various files. [1] [2] [3]