1. Remove checkpoint submodule, replace with downloads from zenodo

[aniongithub]

2. Add map_location so the correct cuda device is used, not the one the model was serialized with
3. Add requirements.txt
4. Add gitignore so eval results and intermediate files aren't added to the index
5. Add dev.Dockerfile to set up all dependencies
6. Add devcontainer.json to set up devcontainer
7. Add launch.json that runs gen_motion_script.py with text prompts from input.txt
8. Remove some tensorboard logging code due to version issues. This shouldn't affect training or inference, only logging. FIX.