Mldock

- fixed a bug on local predict not having --payload argument
- added configure init to aid user in creating .mldock/config
- changed .mldock/config to a workspace level configuration instead of global
- added better error handling for missing project mldock.json and ./mldock/ cli configuration files
- split out decoders/image and other numpy decoders to help with separability in use-cases for developers

- symbolic links don't work with the container does not run in a root environment. As we don't expect ml containers to be open to external traffic, we removed the need to log stderr/stdout to same logs as nginx logs.
- this should allow the containers to truly deploy anywhere.

- adds image/jpeg content-type for `local predict` that expects json response from API
- adds image/jpeg decoder in mldock/inference helpers

- gs asset management: download folder fix + tests
- aws s3 asset managment: tests
- mldock CLI prompt for create project else raise clear error prompt for user