misc cleanup

[christopherhesse]

• Add README.md
• Add a few docstrings
• Add assets to the repo
• Removed generated moc file and IDE files
• Move RandomActionEnv to wrappers. RandomActionWrapper
• Remove unused imports and unused variables
• Change Config to create/use ./saved_models by default
• Change initialize_args to optionally ignore command line arguments so that it's easier to use from a script that doesn't use Config directly (for example, pytest)
• Error if enjoy.py is run without a saved model
• Separate script to run the interactive coinrun environment
• Separate test script that just makes sure you can instantiate the environment at all and take a few steps
• Add some docstrings
• Remove setup_and_load_with_args since it seems to be broken and not used anyway
• Add requirements.txt with dependencies

[christopherhesse]

is_training doesn't appear to be defined here: https://github.com/openai/coinrun/blob/b1e913ae21d550347ca21a7d3a24e0863bbdd55b/coinrun/policies.py#L20